Downstep compound pitch-accent benchmark: 89,539 compounds, listed/unlisted/hard

.gitattributes

@@ -58,3 +58,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 # Video files - compressed
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
+train.jsonl filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,268 @@
+---
+language:
+  - ja
+license: bsd-3-clause
+license_link: https://opensource.org/license/bsd-3-clause
+pretty_name: Downstep compound pitch-accent benchmark (listed / unlisted / hard)
+task_categories:
+  - token-classification
+tags:
+  - japanese
+  - pitch-accent
+  - prosody
+  - tts
+  - openjtalk
+  - compounds
+  - generalization
+  - data-splits
+  - multi-label
+annotations_creators:
+  - machine-generated
+language_creators:
+  - expert-generated
+size_categories:
+  - 10K<n<100K
+configs:
+  - config_name: default
+    data_files:
+      - split: train
+        path: train.jsonl
+      - split: dev
+        path: dev.jsonl
+      - split: test
+        path: test.jsonl
+---
+
+# Downstep compound pitch-accent benchmark
+
+**Japanese noun compounds, their mora segmentation, and every accent their source
+dictionaries attest. Built so that "the model has never seen this compound" is a
+condition you can actually turn on, rather than a claim you have to trust.**
+
+> **No human annotation is present in this release.** Every accent label here is
+> dictionary-derived. `docs/ANNOTATION_GUIDELINES.md` ships the protocol, the CSV
+> format and the agreement statistics for collecting native-speaker judgements on
+> novel compounds — but that collection has not happened, and nothing in this
+> repository should be read as if it had. There are no annotator IDs, no agreement
+> scores and no human-adjudicated items.
+
+---
+
+## Sources and licences
+
+Two independently produced lexicons. Downstep reads the copies installed by
+`unidic-lite` and `pyopenjtalk` and never fetches anything over the network.
+
+| Source | Holder | Licence | Reached via |
+|---|---|---|---|
+| **UniDic 2.1.2** (unidic-mecab) | The UniDic Consortium / NINJAL | GPL **/** LGPL **/** BSD 3-Clause, tri-licensed — **Downstep relies on the BSD option** | `unidic-lite` |
+| **NAIST-jdic** (Open JTalk's system dictionary) | Nara Institute of Science and Technology | BSD 3-Clause | `pyopenjtalk` |
+
+**Attribution both licences require**, reproduced in full in [`NOTICE`](../NOTICE) and
+required of anyone redistributing these rows:
+
+- Retain the copyright notice, the list of BSD 3-Clause conditions and the warranty
+  disclaimer.
+- Do **not** use "The UniDic Consortium", "NINJAL", "Nara Institute of Science and
+  Technology" or their contributors to endorse or promote anything derived from this
+  data without prior written permission. Downstep makes no such endorsement claim.
+
+The BSD option is what makes the derived rows redistributable at all: under the GPL
+option a redistributed derivative would have to be GPL, which is why the choice is
+stated explicitly rather than left to the reader.
+
+The tooling in this repository is Apache-2.0. The data rows below are governed by
+BSD 3-Clause through their sources.
+
+### What is redistributed, and what is not
+
+**Redistributed:** derived rows — surface, reading, mora segmentation, the set of
+attested accents, the component decomposition, and the split/stratum labels.
+
+**Not redistributed:** the dictionaries themselves, in compiled or CSV form; any
+audio; any human judgement. `downstep.lexicon.load_lexicons()` rebuilds the local
+cache (`data/lexicon/*.jsonl`, git-ignored) from the installed packages in ~20 s.
+
+Measured in the cached dump on 2026-09-03 (`make lexicon`, then a count over the
+JSONL):
+
+| | UniDic 2.1.2 | NAIST-jdic |
+|---|---:|---:|
+| entries | 751,705 | 788,331 |
+| distinct surfaces | 568,123 | 461,975 |
+| noun entries | 277,189 | 455,743 |
+| distinct noun surfaces with an accent | 258,371 | 418,671 |
+| noun entries recording **more than one** accent | 21,568 | 0 |
+| noun surfaces whose accents disagree **across** entries | 24,844 | 5,854 |
+
+Those two right-hand columns are the raw material for the multiple-answer field.
+NAIST-jdic records one accent per entry, so its variation only appears when two
+entries share a surface; UniDic records variation inside a single entry
+(`"0,4,3,1"`, CSV-quoted — a naive `split(",")` both mis-counts the fields and
+destroys exactly these rows).
+
+---
+
+## What one row is
+
+A noun compound that satisfies **both** of:
+
+1. It is a lexicalised entry in NAIST-jdic, so a lexicographer assigned it an accent.
+   That is the gold.
+2. It decomposes into **two or more** morphemes whose readings compose back to the
+   compound's own reading, with 連濁 / 促音化 accounted for. That is what makes it a
+   composition problem rather than a lookup.
+
+Cut points for (2) are constrained to a morphological analyser's boundaries. The
+analyser is **UniDic** while the gold is **NAIST-jdic** — deliberately a different
+dictionary, so the decomposition is not read off the answer. Without that constraint
+the decomposition collapses: a naive dynamic program over dictionary surfaces splits
+ファインマン into ファ+イン+マ+ン, because nearly every single kana is itself an entry,
+and reading-consistency alone cannot reject it.
+
+## Fields
+
+| Field | Type | Description |
+|---|---|---|
+| `surface` | str | The compound |
+| `reading` | str | Katakana reading, devoicing marks and interpuncts removed |
+| `moras` | list[str] | Mora segmentation; `len(moras) == n_moras` is enforced at build time |
+| `n_moras` | int | Mora count |
+| `gold_accents` | list[int] | **A set, not a value.** `0` = 平板; `k` = nucleus on mora `k`, 1-indexed |
+| `has_variation` | bool | `len(gold_accents) > 1` — the dictionaries attest more than one accent |
+| `components` | list[obj] | Per component: `surface`, `reading`, `accent`, `n_moras`, `goshu`, `acon_type`, `in_lexicon` |
+| `n_components` | int | Number of components (always ≥ 2) |
+| `goshu` | str | 語種 of the compound, taken as the modal component value (和 / 漢 / 外 / 固 / 混) |
+| `script` | str | `kanji` / `katakana` / `mixed` |
+| `is_proper` | bool | The NAIST entry's POS is 固有名詞 |
+| `all_components_in_lexicon` | bool | Every component was found in a source lexicon |
+| `split` | str | `train` / `dev` / `test` |
+| `hard` | bool | Test-only: **no** component of this compound occurs in any training compound |
+| `source` | str | `naist-jdic`, or `naist-jdic+unidic` when the accent set draws on both |
+
+### `gold_accents` is the multiple-accepted-answers field
+
+Japanese pitch accent genuinely admits more than one correct answer for many words,
+and the dictionaries record that. Scoring against a single "true" accent would
+manufacture errors out of real variation. So:
+
+- **Score membership**, not equality: a prediction is correct if it is *in*
+  `gold_accents`. `downstep.metrics.nucleus_exact_match` takes a set.
+- `has_variation` marks these rows so they can be reported as their own stratum,
+  because a system that always guesses the majority variant looks better on them than
+  it deserves.
+- This field records **lexicographic** variation between and within dictionaries. It
+  is not, and must not be presented as, a survey of how speakers pronounce these
+  words.
+
+---
+
+## Splits, conditions and strata — three different things
+
+**Splits** partition the rows. Assignment is `sha1(surface)` thresholded, not a
+shuffled seed, so adding new candidate words never moves an existing word between
+splits and any machine reproduces the same partition.
+
+| split | share | rows |
+|---|---|---|
+| `train` | 79.9% | **71,544** |
+| `dev` | 10.0% (calibration and abstention threshold) | **8,959** |
+| `test` | 10.1% | **9,036** |
+
+Total: **89,539** compounds.
+
+**Conditions** are not splits. `listed` and `unlisted` are the *same* test rows
+evaluated twice, with the dictionary's whole-word entry present and then ablated at
+query time. See [`docs/METHODOLOGY.md`](../docs/METHODOLOGY.md) for why the ablation
+must be rebuilt per target — batching corrupted 47.7% of targets in the pilot.
+
+**`hard` is a partition of the test set.** Two nested flags:
+
+* `hard` (818 rows, 9.0% of test) — the
+  compound's **final** morpheme never occurs in any training compound. The final element
+  (後部要素) governs compound accent in the classical rules, so this is the linguistically
+  meaningful generalisation test.
+* `hard_all` (105 rows) — *no* component occurs in training. Stricter,
+  but too small to carry a claim; kept as a stratum.
+
+Reported separately, never averaged into the headline: a pilot found a memorisation-shaped
+model gained **nothing** on unseen final components (58.3% vs 58.3%). The trained model does
+better than that — but the split exists so that can be checked rather than assumed.
+
+| stratum field | why it is a stratum |
+|---|---|
+| `is_proper` | Proper nouns are a different problem with different conventions; they are 66% of these rows (59,430 of 89,539) and are **reported on their own**, never folded into a single number |
+| `goshu` | 和語 / 漢語 / 外来語 compounds follow different accent regularities |
+| `n_components` | Two-part compounds are much easier than three-part ones |
+| `n_moras` | Longer compounds have more positions to get wrong |
+| `has_variation` | Rows with more than one accepted answer |
+| `script` | Katakana (largely loanword) compounds behave unlike kanji ones |
+
+Row counts, from `benchmarks/stats.json`:
+
+| stratum | counts |
+|---|---|
+| script | kanji 55,237 · mixed 27,290 · katakana 7,012 |
+| 語種 (`goshu`) | 和 34,760 · 固 25,085 · 漢 23,313 · 外 4,443 · 混 604 · 記号 570 · unknown 764 |
+| components | 2 → 70,419 · 3 → 16,286 · 4 → 2,539 · 5+ → 295 |
+| moras | 4 → 21,228 · 5 → 19,522 · 6 → 18,884 · 7 → 11,112 · 8+ → 18,793 |
+| noun type | proper 59,430 · common 30,109 |
+| more than one attested accent | **1,030** |
+| every component in a lexicon | 89,539 (all of them — decomposition requires it) |
+
+---
+
+## Limitations, stated plainly
+
+- **The gold is a dictionary's opinion.** Dictionaries disagree with each other and
+  with speakers. Where they disagree with each other the row keeps both answers; where
+  they disagree with speakers, this data cannot tell.
+- **The variety recorded is a broadly Tokyo-based standard.** That is what these two
+  lexicons encode, so it is a scoping decision forced by the available data. It is
+  **not** a claim that other regional or generational accent systems are incorrect, and
+  a Downstep prediction is not a correction of anyone's speech.
+- **These are all words that exist in a dictionary.** The target application is
+  compounds that do not. The ablation device is what bridges that gap, and its limits
+  are argued — not waved away — in `docs/METHODOLOGY.md`.
+- **Reading-consistent decomposition rejects real compounds.** The rendaku model covers
+  連濁 and 促音化 but not 連声 or stem alternations (雨 → あま, 酒 → さか), so compounds
+  built on those are dropped rather than mis-segmented. That is a coverage limitation
+  the numbers do not show.
+- **No speech.** No F0 contours, no durations, no speaker information. Accent here is
+  an integer, not a measurement.
+- **Not a pronunciation dictionary.** Do not use these rows as a reference for how a
+  word "should" be said. For that, consult NHK日本語発音アクセント新辞典 or
+  明解日本語アクセント辞典.
+
+## Out-of-scope uses
+
+- As an authority on correct pronunciation, in education or assessment.
+- To evaluate or rank speakers, or to flag any pronunciation as an error.
+- As evidence about regional variation, which it contains no information about.
+
+## Reproducing
+
+```bash
+pip install -e ".[all]"
+make bench        # writes benchmarks/{train,dev,test}.jsonl and benchmarks/stats.json
+```
+
+`scripts/build_bench.py` is the only path from the two dictionaries to these rows,
+and it asserts its own `hard` property rather than trusting it. Nothing is pushed
+anywhere: `--push` exists, defaults off, and refuses to publish on its own.
+
+## Citation
+
+```
+Downstep: a compound pitch-accent benchmark with a dictionary-ablation
+condition for words the TTS dictionary does not have.
+Derived from UniDic 2.1.2 (The UniDic Consortium / NINJAL, BSD-3 option)
+and NAIST-jdic (Nara Institute of Science and Technology, BSD-3).
+No human annotation.
+```
+
+---
+
+*Substantiates: **unlisted-word accuracy** — these rows plus the ablation condition are
+what "the dictionary does not have this compound" is measured on; and **reproducibility**,
+since `make bench` is the only path from the two dictionaries to them.*

baseline_results.json

@@ -0,0 +1,1834 @@
+{
+  "generated_by": "scripts/evaluate.py",
+  "checkpoint": "checkpoints/base",
+  "conditions": {
+    "listed": {
+      "n_items": 9036,
+      "n_reading_matched": 5414,
+      "n_reading_ok_this_condition": 8815,
+      "n_excluded_for_reading": 3622,
+      "openjtalk_seconds": 0.61,
+      "downstep_seconds": 18.13,
+      "strata": {
+        "all": {
+          "n": 5414,
+          "A_openjtalk": 0.9637975618766161,
+          "B_dictionary": 1.0,
+          "A_wrongly_split": 0.022349464351680826,
+          "C_downstep": 0.8134466198743997,
+          "C_wrongly_split": 0.0
+        },
+        "common_nouns": {
+          "n": 2033,
+          "A_openjtalk": 0.9567142154451549,
+          "B_dictionary": 1.0,
+          "A_wrongly_split": 0.017707820954254796,
+          "C_downstep": 0.7968519429414658,
+          "C_wrongly_split": 0.0
+        },
+        "proper_nouns": {
+          "n": 3381,
+          "A_openjtalk": 0.9680567879325643,
+          "B_dictionary": 1.0,
+          "A_wrongly_split": 0.025140490979000295,
+          "C_downstep": 0.8234250221827861,
+          "C_wrongly_split": 0.0
+        },
+        "hard": {
+          "n": 574,
+          "A_openjtalk": 0.9686411149825784,
+          "B_dictionary": 1.0,
+          "A_wrongly_split": 0.0156794425087108,
+          "C_downstep": 0.7334494773519163,
+          "C_wrongly_split": 0.0
+        },
+        "hard_all": {
+          "n": 98,
+          "A_openjtalk": 0.9897959183673469,
+          "B_dictionary": 1.0,
+          "A_wrongly_split": 0.0,
+          "C_downstep": 0.6224489795918368,
+          "C_wrongly_split": 0.0
+        },
+        "variation": {
+          "n": 49,
+          "A_openjtalk": 0.9387755102040817,
+          "B_dictionary": 1.0,
+          "A_wrongly_split": 0.061224489795918366,
+          "C_downstep": 0.6122448979591837,
+          "C_wrongly_split": 0.0
+        },
+        "goshu_和": {
+          "n": 1988,
+          "A_openjtalk": 0.954225352112676,
+          "B_dictionary": 1.0,
+          "A_wrongly_split": 0.02665995975855131,
+          "C_downstep": 0.7907444668008048,
+          "C_wrongly_split": 0.0
+        },
+        "goshu_漢": {
+          "n": 1349,
+          "A_openjtalk": 0.9747961452928094,
+          "B_dictionary": 1.0,
+          "A_wrongly_split": 0.015567086730911787,
+          "C_downstep": 0.8413639733135656,
+          "C_wrongly_split": 0.0
+        },
+        "goshu_外": {
+          "n": 405,
+          "A_openjtalk": 0.9481481481481482,
+          "B_dictionary": 1.0,
+          "A_wrongly_split": 0.024691358024691357,
+          "C_downstep": 0.8222222222222222,
+          "C_wrongly_split": 0.0
+        },
+        "goshu_混": {
+          "n": 41,
+          "A_openjtalk": 1.0,
+          "B_dictionary": 1.0,
+          "A_wrongly_split": 0.0,
+          "C_downstep": 0.8780487804878049,
+          "C_wrongly_split": 0.0
+        },
+        "goshu_固": {
+          "n": 1562,
+          "A_openjtalk": 0.969270166453265,
+          "B_dictionary": 1.0,
+          "A_wrongly_split": 0.023687580025608196,
+          "C_downstep": 0.8149807938540333,
+          "C_wrongly_split": 0.0
+        },
+        "script_kanji": {
+          "n": 2884,
+          "A_openjtalk": 0.9680998613037448,
+          "B_dictionary": 1.0,
+          "A_wrongly_split": 0.025312066574202496,
+          "C_downstep": 0.8311373092926491,
+          "C_wrongly_split": 0.0
+        },
+        "script_katakana": {
+          "n": 683,
+          "A_openjtalk": 0.95900439238653,
+          "B_dictionary": 1.0,
+          "A_wrongly_split": 0.020497803806734993,
+          "C_downstep": 0.7862371888726208,
+          "C_wrongly_split": 0.0
+        },
+        "script_mixed": {
+          "n": 1847,
+          "A_openjtalk": 0.9588521927449919,
+          "B_dictionary": 1.0,
+          "A_wrongly_split": 0.018408229561451002,
+          "C_downstep": 0.7958852192744992,
+          "C_wrongly_split": 0.0
+        },
+        "components_2": {
+          "n": 4595,
+          "A_openjtalk": 0.9608269858541894,
+          "B_dictionary": 1.0,
+          "A_wrongly_split": 0.02372143634385201,
+          "C_downstep": 0.8187159956474429,
+          "C_wrongly_split": 0.0
+        },
+        "components_3": {
+          "n": 735,
+          "A_openjtalk": 0.9782312925170068,
+          "B_dictionary": 1.0,
+          "A_wrongly_split": 0.0163265306122449,
+          "C_downstep": 0.7850340136054422,
+          "C_wrongly_split": 0.0
+        },
+        "components_4": {
+          "n": 80,
+          "A_openjtalk": 1.0,
+          "B_dictionary": 1.0,
+          "A_wrongly_split": 0.0,
+          "C_downstep": 0.775,
+          "C_wrongly_split": 0.0
+        }
+      },
+      "abstention": {
+        "curve": [
+          {
+            "threshold": 0.287819,
+            "coverage": 0.999815293683044,
+            "combined_accuracy": 0.8136313261913557,
+            "covered_accuracy": 0.8135968963606134,
+            "regressions": 955
+          },
+          {
+            "threshold": 0.404532,
+            "coverage": 0.9898411525674178,
+            "combined_accuracy": 0.819541928333949,
+            "covered_accuracy": 0.8186228774025005,
+            "regressions": 921
+          },
+          {
+            "threshold": 0.453785,
+            "coverage": 0.9798670114517917,
+            "combined_accuracy": 0.8254525304765423,
+            "covered_accuracy": 0.8233741753063148,
+            "regressions": 888
+          },
+          {
+            "threshold": 0.48362,
+            "coverage": 0.9698928703361654,
+            "combined_accuracy": 0.8313631326191355,
+            "covered_accuracy": 0.8276518758331747,
+            "regressions": 856
+          },
+          {
+            "threshold": 0.50118,
+            "coverage": 0.9599187292205393,
+            "combined_accuracy": 0.8370890284447728,
+            "covered_accuracy": 0.8322108908985953,
+            "regressions": 824
+          },
+          {
+            "threshold": 0.518026,
+            "coverage": 0.9499445881049132,
+            "combined_accuracy": 0.842260805319542,
+            "covered_accuracy": 0.8362823254909586,
+            "regressions": 795
+          },
+          {
+            "threshold": 0.53355,
+            "coverage": 0.939970446989287,
+            "combined_accuracy": 0.8465090506095309,
+            "covered_accuracy": 0.8400471605423462,
+            "regressions": 771
+          },
+          {
+            "threshold": 0.550858,
+            "coverage": 0.9299963058736609,
+            "combined_accuracy": 0.851496121167344,
+            "covered_accuracy": 0.8438927507447865,
+            "regressions": 743
+          },
+          {
+            "threshold": 0.568295,
+            "coverage": 0.9198374584410787,
+            "combined_accuracy": 0.856483191725157,
+            "covered_accuracy": 0.8481927710843373,
+            "regressions": 714
+          },
+          {
+            "threshold": 0.5827,
+            "coverage": 0.9098633173254526,
+            "combined_accuracy": 0.8598079054303657,
+            "covered_accuracy": 0.8509947218838815,
+            "regressions": 694
+          },
+          {
+            "threshold": 0.595637,
+            "coverage": 0.8998891762098263,
+            "combined_accuracy": 0.8642408570373107,
+            "covered_accuracy": 0.8548850574712644,
+            "regressions": 670
+          },
+          {
+            "threshold": 0.611318,
+            "coverage": 0.8899150350942002,
+            "combined_accuracy": 0.8690432212781677,
+            "covered_accuracy": 0.8590701535907015,
+            "regressions": 643
+          },
+          {
+            "threshold": 0.626212,
+            "coverage": 0.8799408939785741,
+            "combined_accuracy": 0.8723679349833764,
+            "covered_accuracy": 0.8618807724601175,
+            "regressions": 624
+          },
+          {
+            "threshold": 0.640905,
+            "coverage": 0.8699667528629479,
+            "combined_accuracy": 0.8762467676394533,
+            "covered_accuracy": 0.8651804670912951,
+            "regressions": 601
+          },
+          {
+            "threshold": 0.655575,
+            "coverage": 0.8599926117473218,
+            "combined_accuracy": 0.879571481344662,
+            "covered_accuracy": 0.8679123711340206,
+            "regressions": 581
+          },
+          {
+            "threshold": 0.672725,
+            "coverage": 0.8500184706316956,
+            "combined_accuracy": 0.8836350203176949,
+            "covered_accuracy": 0.8720121686223381,
+            "regressions": 556
+          },
+          {
+            "threshold": 0.686029,
+            "coverage": 0.8398596231991134,
+            "combined_accuracy": 0.8869597340229036,
+            "covered_accuracy": 0.8744226962832637,
+            "regressions": 538
+          },
+          {
+            "threshold": 0.700345,
+            "coverage": 0.8298854820834872,
+            "combined_accuracy": 0.8900997414111562,
+            "covered_accuracy": 0.8773647896728244,
+            "regressions": 519
+          },
+          {
+            "threshold": 0.712894,
+            "coverage": 0.8199113409678611,
+            "combined_accuracy": 0.8932397487994089,
+            "covered_accuracy": 0.8806037395809867,
+            "regressions": 500
+          },
+          {
+            "threshold": 0.72452,
+            "coverage": 0.809937199852235,
+            "combined_accuracy": 0.8960103435537495,
+            "covered_accuracy": 0.883466362599772,
+            "regressions": 482
+          },
+          {
+            "threshold": 0.73972,
+            "coverage": 0.7999630587366088,
+            "combined_accuracy": 0.8978574067233099,
+            "covered_accuracy": 0.8857076887554837,
+            "regressions": 468
+          },
+          {
+            "threshold": 0.754053,
+            "coverage": 0.7899889176209827,
+            "combined_accuracy": 0.9017362393793867,
+            "covered_accuracy": 0.8891746551321019,
+            "regressions": 447
+          },
+          {
+            "threshold": 0.765463,
+            "coverage": 0.7800147765053564,
+            "combined_accuracy": 0.9052456594015516,
+            "covered_accuracy": 0.8924934880416765,
+            "regressions": 428
+          },
+          {
+            "threshold": 0.77653,
+            "coverage": 0.7700406353897303,
+            "combined_accuracy": 0.9065386036202439,
+            "covered_accuracy": 0.893739505876709,
+            "regressions": 418
+          },
+          {
+            "threshold": 0.785172,
+            "coverage": 0.7598817879571481,
+            "combined_accuracy": 0.9089397857406724,
+            "covered_accuracy": 0.8964511424404472,
+            "regressions": 403
+          },
+          {
+            "threshold": 0.794848,
+            "coverage": 0.749907646841522,
+            "combined_accuracy": 0.9102327299593647,
+            "covered_accuracy": 0.8972906403940887,
+            "regressions": 396
+          },
+          {
+            "threshold": 0.804995,
+            "coverage": 0.7399335057258958,
+            "combined_accuracy": 0.9124492057628371,
+            "covered_accuracy": 0.8996505242136795,
+            "regressions": 381
+          },
+          {
+            "threshold": 0.817838,
+            "coverage": 0.7299593646102697,
+            "combined_accuracy": 0.9142962689323975,
+            "covered_accuracy": 0.9015688259109311,
+            "regressions": 369
+          },
+          {
+            "threshold": 0.826393,
+            "coverage": 0.7199852234946436,
+            "combined_accuracy": 0.9157739194680458,
+            "covered_accuracy": 0.9030271934325295,
+            "regressions": 360
+          },
+          {
+            "threshold": 0.835651,
+            "coverage": 0.7100110823790173,
+            "combined_accuracy": 0.916697451052826,
+            "covered_accuracy": 0.9037460978147763,
+            "regressions": 352
+          },
+          {
+            "threshold": 0.8443,
+            "coverage": 0.7000369412633912,
+            "combined_accuracy": 0.9183598079054304,
+            "covered_accuracy": 0.9055408970976253,
+            "regressions": 341
+          },
+          {
+            "threshold": 0.853646,
+            "coverage": 0.689878093830809,
+            "combined_accuracy": 0.9203915773919468,
+            "covered_accuracy": 0.9081659973226238,
+            "regressions": 327
+          },
+          {
+            "threshold": 0.861325,
+            "coverage": 0.6799039527151829,
+            "combined_accuracy": 0.9216845216106391,
+            "covered_accuracy": 0.9092637870143983,
+            "regressions": 318
+          },
+          {
+            "threshold": 0.868863,
+            "coverage": 0.6699298115995567,
+            "combined_accuracy": 0.9229774658293314,
+            "covered_accuracy": 0.9098428453267163,
+            "regressions": 311
+          },
+          {
+            "threshold": 0.877202,
+            "coverage": 0.6599556704839306,
+            "combined_accuracy": 0.925932766900628,
+            "covered_accuracy": 0.9129582983487265,
+            "regressions": 295
+          },
+          {
+            "threshold": 0.883933,
+            "coverage": 0.6499815293683044,
+            "combined_accuracy": 0.9274104174362763,
+            "covered_accuracy": 0.9147485080988917,
+            "regressions": 284
+          },
+          {
+            "threshold": 0.890918,
+            "coverage": 0.6400073882526782,
+            "combined_accuracy": 0.9287033616549686,
+            "covered_accuracy": 0.9163059163059163,
+            "regressions": 274
+          },
+          {
+            "threshold": 0.895454,
+            "coverage": 0.6300332471370521,
+            "combined_accuracy": 0.930735131141485,
+            "covered_accuracy": 0.9187921430665494,
+            "regressions": 261
+          },
+          {
+            "threshold": 0.900932,
+            "coverage": 0.6200591060214259,
+            "combined_accuracy": 0.9314739564093092,
+            "covered_accuracy": 0.9192731605600238,
+            "regressions": 256
+          },
+          {
+            "threshold": 0.905733,
+            "coverage": 0.6099002585888438,
+            "combined_accuracy": 0.9335057258958256,
+            "covered_accuracy": 0.922168382798304,
+            "regressions": 242
+          },
+          {
+            "threshold": 0.911608,
+            "coverage": 0.5999261174732176,
+            "combined_accuracy": 0.935352789065386,
+            "covered_accuracy": 0.9245689655172413,
+            "regressions": 232
+          },
+          {
+            "threshold": 0.916809,
+            "coverage": 0.5899519763575914,
+            "combined_accuracy": 0.9368304396010344,
+            "covered_accuracy": 0.926111458985598,
+            "regressions": 224
+          },
+          {
+            "threshold": 0.920427,
+            "coverage": 0.5799778352419652,
+            "combined_accuracy": 0.9379386775027706,
+            "covered_accuracy": 0.9273885350318471,
+            "regressions": 216
+          },
+          {
+            "threshold": 0.925197,
+            "coverage": 0.5700036941263391,
+            "combined_accuracy": 0.9405245659401551,
+            "covered_accuracy": 0.9309786130913804,
+            "regressions": 201
+          },
+          {
+            "threshold": 0.929965,
+            "coverage": 0.560029553010713,
+            "combined_accuracy": 0.9421869227927595,
+            "covered_accuracy": 0.933377308707124,
+            "regressions": 191
+          },
+          {
+            "threshold": 0.933174,
+            "coverage": 0.5500554118950868,
+            "combined_accuracy": 0.9438492796453639,
+            "covered_accuracy": 0.9358629952988583,
+            "regressions": 180
+          },
+          {
+            "threshold": 0.93709,
+            "coverage": 0.5400812707794607,
+            "combined_accuracy": 0.9447728112301441,
+            "covered_accuracy": 0.9374145006839946,
+            "regressions": 174
+          },
+          {
+            "threshold": 0.940724,
+            "coverage": 0.5299224233468784,
+            "combined_accuracy": 0.9458810491318803,
+            "covered_accuracy": 0.9393516904844894,
+            "regressions": 165
+          },
+          {
+            "threshold": 0.944268,
+            "coverage": 0.5199482822312523,
+            "combined_accuracy": 0.9466198743997045,
+            "covered_accuracy": 0.9399644760213144,
+            "regressions": 161
+          },
+          {
+            "threshold": 0.948099,
+            "coverage": 0.5099741411156261,
+            "combined_accuracy": 0.9471739933505726,
+            "covered_accuracy": 0.9406012314378849,
+            "regressions": 156
+          },
+          {
+            "threshold": 0.951436,
+            "coverage": 0.5,
+            "combined_accuracy": 0.9475434059844847,
+            "covered_accuracy": 0.9408939785740672,
+            "regressions": 152
+          },
+          {
+            "threshold": 0.954023,
+            "coverage": 0.49002585888437383,
+            "combined_accuracy": 0.9495751754710011,
+            "covered_accuracy": 0.944214097248398,
+            "regressions": 140
+          },
+          {
+            "threshold": 0.956498,
+            "coverage": 0.48005171776874767,
+            "combined_accuracy": 0.9504987070557813,
+            "covered_accuracy": 0.9453636013851482,
+            "regressions": 134
+          },
+          {
+            "threshold": 0.958537,
+            "coverage": 0.47007757665312155,
+            "combined_accuracy": 0.9514222386405615,
+            "covered_accuracy": 0.9469548133595285,
+            "regressions": 127
+          },
+          {
+            "threshold": 0.960376,
+            "coverage": 0.45991872922053934,
+            "combined_accuracy": 0.9521610639083856,
+            "covered_accuracy": 0.948995983935743,
+            "regressions": 120
+          },
+          {
+            "threshold": 0.963265,
+            "coverage": 0.44994458810491317,
+            "combined_accuracy": 0.9530845954931658,
+            "covered_accuracy": 0.9499178981937603,
+            "regressions": 115
+          },
+          {
+            "threshold": 0.965379,
+            "coverage": 0.43997044698928706,
+            "combined_accuracy": 0.9532693018101219,
+            "covered_accuracy": 0.9500419815281276,
+            "regressions": 113
+          },
+          {
+            "threshold": 0.96692,
+            "coverage": 0.4299963058736609,
+            "combined_accuracy": 0.9541928333949021,
+            "covered_accuracy": 0.9510309278350515,
+            "regressions": 108
+          },
+          {
+            "threshold": 0.968873,
+            "coverage": 0.4200221647580347,
+            "combined_accuracy": 0.9549316586627262,
+            "covered_accuracy": 0.951627088830255,
+            "regressions": 104
+          },
+          {
+            "threshold": 0.970763,
+            "coverage": 0.41004802364240855,
+            "combined_accuracy": 0.9554857776135944,
+            "covered_accuracy": 0.9518018018018019,
+            "regressions": 101
+          },
+          {
+            "threshold": 0.972469,
+            "coverage": 0.40007388252678244,
+            "combined_accuracy": 0.9558551902475064,
+            "covered_accuracy": 0.9519852262234534,
+            "regressions": 98
+          },
+          {
+            "threshold": 0.974047,
+            "coverage": 0.3900997414111563,
+            "combined_accuracy": 0.9562246028814185,
+            "covered_accuracy": 0.9535984848484849,
+            "regressions": 94
+          },
+          {
+            "threshold": 0.975516,
+            "coverage": 0.37994089397857406,
+            "combined_accuracy": 0.9564093091983746,
+            "covered_accuracy": 0.9533300923675255,
+            "regressions": 92
+          },
+          {
+            "threshold": 0.976913,
+            "coverage": 0.3699667528629479,
+            "combined_accuracy": 0.9571481344661987,
+            "covered_accuracy": 0.9550673989016475,
+            "regressions": 86
+          },
+          {
+            "threshold": 0.978492,
+            "coverage": 0.3599926117473218,
+            "combined_accuracy": 0.9577022534170668,
+            "covered_accuracy": 0.9563878912262699,
+            "regressions": 82
+          },
+          {
+            "threshold": 0.979541,
+            "coverage": 0.3500184706316956,
+            "combined_accuracy": 0.9573328407831548,
+            "covered_accuracy": 0.9551451187335093,
+            "regressions": 82
+          },
+          {
+            "threshold": 0.980824,
+            "coverage": 0.34004432951606944,
+            "combined_accuracy": 0.9578869597340229,
+            "covered_accuracy": 0.955458989679522,
+            "regressions": 79
+          },
+          {
+            "threshold": 0.981792,
+            "coverage": 0.3300701884004433,
+            "combined_accuracy": 0.9580716660509789,
+            "covered_accuracy": 0.9557918298824846,
+            "regressions": 77
+          },
+          {
+            "threshold": 0.98304,
+            "coverage": 0.32009604728481716,
+            "combined_accuracy": 0.958256372367935,
+            "covered_accuracy": 0.9561454125793422,
+            "regressions": 74
+          },
+          {
+            "threshold": 0.984012,
+            "coverage": 0.30993719985223495,
+            "combined_accuracy": 0.9591799039527151,
+            "covered_accuracy": 0.9576877234803337,
+            "regressions": 69
+          },
+          {
+            "threshold": 0.984861,
+            "coverage": 0.2999630587366088,
+            "combined_accuracy": 0.9593646102696712,
+            "covered_accuracy": 0.9568965517241379,
+            "regressions": 68
+          },
+          {
+            "threshold": 0.985914,
+            "coverage": 0.2899889176209826,
+            "combined_accuracy": 0.9593646102696712,
+            "covered_accuracy": 0.9573248407643312,
+            "regressions": 66
+          },
+          {
+            "threshold": 0.986784,
+            "coverage": 0.2800147765053565,
+            "combined_accuracy": 0.9602881418544514,
+            "covered_accuracy": 0.9597625329815304,
+            "regressions": 60
+          },
+          {
+            "threshold": 0.98768,
+            "coverage": 0.27004063538973033,
+            "combined_accuracy": 0.9601034355374953,
+            "covered_accuracy": 0.9596443228454172,
+            "regressions": 58
+          },
+          {
+            "threshold": 0.988424,
+            "coverage": 0.26006649427410417,
+            "combined_accuracy": 0.9612116734392316,
+            "covered_accuracy": 0.9630681818181818,
+            "regressions": 51
+          },
+          {
+            "threshold": 0.989225,
+            "coverage": 0.250092353158478,
+            "combined_accuracy": 0.9623199113409678,
+            "covered_accuracy": 0.96602658788774,
+            "regressions": 45
+          },
+          {
+            "threshold": 0.989847,
+            "coverage": 0.24011821204285186,
+            "combined_accuracy": 0.9626893239748799,
+            "covered_accuracy": 0.9661538461538461,
+            "regressions": 43
+          },
+          {
+            "threshold": 0.990648,
+            "coverage": 0.22995936461026967,
+            "combined_accuracy": 0.9625046176579239,
+            "covered_accuracy": 0.9654618473895582,
+            "regressions": 42
+          },
+          {
+            "threshold": 0.991326,
+            "coverage": 0.21998522349464353,
+            "combined_accuracy": 0.9625046176579239,
+            "covered_accuracy": 0.964735516372796,
+            "regressions": 41
+          },
+          {
+            "threshold": 0.991868,
+            "coverage": 0.21001108237901736,
+            "combined_accuracy": 0.9626893239748799,
+            "covered_accuracy": 0.9665787159190853,
+            "regressions": 37
+          },
+          {
+            "threshold": 0.992456,
+            "coverage": 0.20003694126339122,
+            "combined_accuracy": 0.963243442925748,
+            "covered_accuracy": 0.9676823638042474,
+            "regressions": 34
+          },
+          {
+            "threshold": 0.993025,
+            "coverage": 0.19006280014776505,
+            "combined_accuracy": 0.963058736608792,
+            "covered_accuracy": 0.966958211856171,
+            "regressions": 33
+          },
+          {
+            "threshold": 0.993511,
+            "coverage": 0.1800886590321389,
+            "combined_accuracy": 0.9626893239748799,
+            "covered_accuracy": 0.9661538461538461,
+            "regressions": 32
+          },
+          {
+            "threshold": 0.993984,
+            "coverage": 0.17011451791651275,
+            "combined_accuracy": 0.963058736608792,
+            "covered_accuracy": 0.9674267100977199,
+            "regressions": 29
+          },
+          {
+            "threshold": 0.99437,
+            "coverage": 0.15995567048393056,
+            "combined_accuracy": 0.963243442925748,
+            "covered_accuracy": 0.9665127020785219,
+            "regressions": 28
+          },
+          {
+            "threshold": 0.994859,
+            "coverage": 0.15016623568526044,
+            "combined_accuracy": 0.9636128555596601,
+            "covered_accuracy": 0.9692496924969249,
+            "regressions": 24
+          },
+          {
+            "threshold": 0.995294,
+            "coverage": 0.14000738825267825,
+            "combined_accuracy": 0.9639822681935722,
+            "covered_accuracy": 0.9709762532981531,
+            "regressions": 22
+          },
+          {
+            "threshold": 0.995697,
+            "coverage": 0.13021795345400813,
+            "combined_accuracy": 0.9645363871444403,
+            "covered_accuracy": 0.9730496453900709,
+            "regressions": 19
+          },
+          {
+            "threshold": 0.996291,
+            "coverage": 0.12005910602142593,
+            "combined_accuracy": 0.9649057997783524,
+            "covered_accuracy": 0.9738461538461538,
+            "regressions": 17
+          },
+          {
+            "threshold": 0.996686,
+            "coverage": 0.11008496490579978,
+            "combined_accuracy": 0.9643516808274843,
+            "covered_accuracy": 0.9714765100671141,
+            "regressions": 17
+          },
+          {
+            "threshold": 0.99705,
+            "coverage": 0.10011082379017362,
+            "combined_accuracy": 0.9645363871444403,
+            "covered_accuracy": 0.974169741697417,
+            "regressions": 14
+          },
+          {
+            "threshold": 0.997386,
+            "coverage": 0.09013668267454747,
+            "combined_accuracy": 0.9643516808274843,
+            "covered_accuracy": 0.9774590163934426,
+            "regressions": 11
+          },
+          {
+            "threshold": 0.997703,
+            "coverage": 0.08016254155892132,
+            "combined_accuracy": 0.9641669745105282,
+            "covered_accuracy": 0.9746543778801844,
+            "regressions": 11
+          },
+          {
+            "threshold": 0.997964,
+            "coverage": 0.07000369412633912,
+            "combined_accuracy": 0.9645363871444403,
+            "covered_accuracy": 0.978891820580475,
+            "regressions": 8
+          },
+          {
+            "threshold": 0.998248,
+            "coverage": 0.060029553010712965,
+            "combined_accuracy": 0.9652752124122645,
+            "covered_accuracy": 0.9907692307692307,
+            "regressions": 3
+          },
+          {
+            "threshold": 0.998595,
+            "coverage": 0.05005541189508681,
+            "combined_accuracy": 0.9649057997783524,
+            "covered_accuracy": 0.988929889298893,
+            "regressions": 3
+          },
+          {
+            "threshold": 0.998848,
+            "coverage": 0.04008127077946066,
+            "combined_accuracy": 0.9647210934613963,
+            "covered_accuracy": 0.9953917050691244,
+            "regressions": 1
+          },
+          {
+            "threshold": 0.99905,
+            "coverage": 0.029922423346878464,
+            "combined_accuracy": 0.9643516808274843,
+            "covered_accuracy": 0.9938271604938271,
+            "regressions": 1
+          },
+          {
+            "threshold": 0.99924,
+            "coverage": 0.020132988548208348,
+            "combined_accuracy": 0.9641669745105282,
+            "covered_accuracy": 0.9908256880733946,
+            "regressions": 1
+          },
+          {
+            "threshold": 0.999454,
+            "coverage": 0.009974141115626155,
+            "combined_accuracy": 0.9643516808274843,
+            "covered_accuracy": 1.0,
+            "regressions": 0
+          },
+          {
+            "threshold": 0.99978,
+            "coverage": 0.0,
+            "combined_accuracy": 0.9637975618766161,
+            "covered_accuracy": null,
+            "regressions": 0
+          }
+        ],
+        "baseline_accuracy": 0.9637975618766161,
+        "no_abstention_accuracy": 0.8134466198743997,
+        "best_combined": {
+          "threshold": 0.998248,
+          "coverage": 0.060029553010712965,
+          "combined_accuracy": 0.9652752124122645,
+          "covered_accuracy": 0.9907692307692307,
+          "regressions": 3
+        },
+        "conservative_zero_regression": {
+          "threshold": 0.999454,
+          "coverage": 0.009974141115626155,
+          "combined_accuracy": 0.9643516808274843,
+          "covered_accuracy": 1.0,
+          "regressions": 0
+        },
+        "ece": 0.0576062929816676
+      },
+      "nbest_coverage": {
+        "all": {
+          "1": 0.8134466198743997,
+          "2": 0.9436645733284078,
+          "3": 0.9804211304026598,
+          "4": 0.9918729220539343,
+          "5": 0.9970446989287034
+        },
+        "variation_only": {
+          "1": 0.6122448979591837,
+          "2": 0.9387755102040817,
+          "3": 0.9795918367346939,
+          "4": 1.0,
+          "5": 1.0
+        },
+        "n_variation": 49
+      },
+      "operating_point_dev_chosen": {
+        "chosen_on": "dev split, same condition",
+        "max_accuracy": {
+          "threshold": 0.994487,
+          "coverage": 0.15755448836350203,
+          "combined_accuracy": 0.963243442925748,
+          "regressions": 27,
+          "improvements": 24,
+          "n": 5414
+        },
+        "zero_regression_on_dev": {
+          "threshold": 0.999395,
+          "coverage": 0.013668267454746953,
+          "combined_accuracy": 0.9641669745105282,
+          "regressions": 1,
+          "improvements": 3,
+          "n": 5414
+        },
+        "dev_zero_regression_coverage": 0.015083798882681564
+      }
+    },
+    "unlisted": {
+      "n_items": 9036,
+      "n_reading_matched": 5414,
+      "n_reading_ok_this_condition": 5462,
+      "n_excluded_for_reading": 3622,
+      "openjtalk_seconds": 4.07,
+      "downstep_seconds": 18.16,
+      "strata": {
+        "all": {
+          "n": 5414,
+          "A_openjtalk": 0.4059844846693757,
+          "B_dictionary": 0.4059844846693757,
+          "A_wrongly_split": 0.2332840783154784,
+          "C_downstep": 0.8134466198743997,
+          "C_wrongly_split": 0.0
+        },
+        "common_nouns": {
+          "n": 2033,
+          "A_openjtalk": 0.31431382193802265,
+          "B_dictionary": 0.31431382193802265,
+          "A_wrongly_split": 0.2867683226758485,
+          "C_downstep": 0.7968519429414658,
+          "C_wrongly_split": 0.0
+        },
+        "proper_nouns": {
+          "n": 3381,
+          "A_openjtalk": 0.46110618160307604,
+          "B_dictionary": 0.46110618160307604,
+          "A_wrongly_split": 0.20112392783200236,
+          "C_downstep": 0.8234250221827861,
+          "C_wrongly_split": 0.0
+        },
+        "hard": {
+          "n": 574,
+          "A_openjtalk": 0.4094076655052265,
+          "B_dictionary": 0.4094076655052265,
+          "A_wrongly_split": 0.21602787456445993,
+          "C_downstep": 0.7334494773519163,
+          "C_wrongly_split": 0.0
+        },
+        "hard_all": {
+          "n": 98,
+          "A_openjtalk": 0.3163265306122449,
+          "B_dictionary": 0.3163265306122449,
+          "A_wrongly_split": 0.19387755102040816,
+          "C_downstep": 0.6224489795918368,
+          "C_wrongly_split": 0.0
+        },
+        "variation": {
+          "n": 49,
+          "A_openjtalk": 0.40816326530612246,
+          "B_dictionary": 0.40816326530612246,
+          "A_wrongly_split": 0.22448979591836735,
+          "C_downstep": 0.6122448979591837,
+          "C_wrongly_split": 0.0
+        },
+        "goshu_和": {
+          "n": 1988,
+          "A_openjtalk": 0.3631790744466801,
+          "B_dictionary": 0.3631790744466801,
+          "A_wrongly_split": 0.23943661971830985,
+          "C_downstep": 0.7907444668008048,
+          "C_wrongly_split": 0.0
+        },
+        "goshu_漢": {
+          "n": 1349,
+          "A_openjtalk": 0.42031134173461826,
+          "B_dictionary": 0.42031134173461826,
+          "A_wrongly_split": 0.21793921423276502,
+          "C_downstep": 0.8413639733135656,
+          "C_wrongly_split": 0.0
+        },
+        "goshu_外": {
+          "n": 405,
+          "A_openjtalk": 0.5506172839506173,
+          "B_dictionary": 0.5506172839506173,
+          "A_wrongly_split": 0.14814814814814814,
+          "C_downstep": 0.8222222222222222,
+          "C_wrongly_split": 0.0
+        },
+        "goshu_混": {
+          "n": 41,
+          "A_openjtalk": 0.24390243902439024,
+          "B_dictionary": 0.24390243902439024,
+          "A_wrongly_split": 0.24390243902439024,
+          "C_downstep": 0.8780487804878049,
+          "C_wrongly_split": 0.0
+        },
+        "goshu_固": {
+          "n": 1562,
+          "A_openjtalk": 0.41037131882202305,
+          "B_dictionary": 0.41037131882202305,
+          "A_wrongly_split": 0.26568501920614596,
+          "C_downstep": 0.8149807938540333,
+          "C_wrongly_split": 0.0
+        },
+        "script_kanji": {
+          "n": 2884,
+          "A_openjtalk": 0.4791955617198336,
+          "B_dictionary": 0.4791955617198336,
+          "A_wrongly_split": 0.17094313453536755,
+          "C_downstep": 0.8311373092926491,
+          "C_wrongly_split": 0.0
+        },
+        "script_katakana": {
+          "n": 683,
+          "A_openjtalk": 0.424597364568082,
+          "B_dictionary": 0.424597364568082,
+          "A_wrongly_split": 0.22547584187408493,
+          "C_downstep": 0.7862371888726208,
+          "C_wrongly_split": 0.0
+        },
+        "script_mixed": {
+          "n": 1847,
+          "A_openjtalk": 0.28478613968597727,
+          "B_dictionary": 0.28478613968597727,
+          "A_wrongly_split": 0.3335138061721711,
+          "C_downstep": 0.7958852192744992,
+          "C_wrongly_split": 0.0
+        },
+        "components_2": {
+          "n": 4595,
+          "A_openjtalk": 0.41653971708378673,
+          "B_dictionary": 0.41653971708378673,
+          "A_wrongly_split": 0.2132752992383025,
+          "C_downstep": 0.8187159956474429,
+          "C_wrongly_split": 0.0
+        },
+        "components_3": {
+          "n": 735,
+          "A_openjtalk": 0.363265306122449,
+          "B_dictionary": 0.363265306122449,
+          "A_wrongly_split": 0.3197278911564626,
+          "C_downstep": 0.7850340136054422,
+          "C_wrongly_split": 0.0
+        },
+        "components_4": {
+          "n": 80,
+          "A_openjtalk": 0.2125,
+          "B_dictionary": 0.2125,
+          "A_wrongly_split": 0.55,
+          "C_downstep": 0.775,
+          "C_wrongly_split": 0.0
+        }
+      },
+      "abstention": {
+        "curve": [
+          {
+            "threshold": 0.287819,
+            "coverage": 0.999815293683044,
+            "combined_accuracy": 0.8134466198743997,
+            "covered_accuracy": 0.8135968963606134,
+            "regressions": 256
+          },
+          {
+            "threshold": 0.404532,
+            "coverage": 0.9898411525674178,
+            "combined_accuracy": 0.8127077946065755,
+            "covered_accuracy": 0.8186228774025005,
+            "regressions": 248
+          },
+          {
+            "threshold": 0.453785,
+            "coverage": 0.9798670114517917,
+            "combined_accuracy": 0.8104913188031031,
+            "covered_accuracy": 0.8233741753063148,
+            "regressions": 244
+          },
+          {
+            "threshold": 0.48362,
+            "coverage": 0.9698928703361654,
+            "combined_accuracy": 0.8090136682674548,
+            "covered_accuracy": 0.8276518758331747,
+            "regressions": 235
+          },
+          {
+            "threshold": 0.50118,
+            "coverage": 0.9599187292205393,
+            "combined_accuracy": 0.8084595493165866,
+            "covered_accuracy": 0.8322108908985953,
+            "regressions": 221
+          },
+          {
+            "threshold": 0.518026,
+            "coverage": 0.9499445881049132,
+            "combined_accuracy": 0.8064277798300702,
+            "covered_accuracy": 0.8362823254909586,
+            "regressions": 215
+          },
+          {
+            "threshold": 0.53355,
+            "coverage": 0.939970446989287,
+            "combined_accuracy": 0.8051348356113779,
+            "covered_accuracy": 0.8400471605423462,
+            "regressions": 209
+          },
+          {
+            "threshold": 0.550858,
+            "coverage": 0.9299963058736609,
+            "combined_accuracy": 0.8040265977096417,
+            "covered_accuracy": 0.8438927507447865,
+            "regressions": 197
+          },
+          {
+            "threshold": 0.568295,
+            "coverage": 0.9198374584410787,
+            "combined_accuracy": 0.8018101219061692,
+            "covered_accuracy": 0.8481927710843373,
+            "regressions": 191
+          },
+          {
+            "threshold": 0.5827,
+            "coverage": 0.9098633173254526,
+            "combined_accuracy": 0.7984854082009605,
+            "covered_accuracy": 0.8509947218838815,
+            "regressions": 184
+          },
+          {
+            "threshold": 0.595637,
+            "coverage": 0.8998891762098263,
+            "combined_accuracy": 0.7964536387144441,
+            "covered_accuracy": 0.8548850574712644,
+            "regressions": 178
+          },
+          {
+            "threshold": 0.611318,
+            "coverage": 0.8899150350942002,
+            "combined_accuracy": 0.7953454008127078,
+            "covered_accuracy": 0.8590701535907015,
+            "regressions": 168
+          },
+          {
+            "threshold": 0.626212,
+            "coverage": 0.8799408939785741,
+            "combined_accuracy": 0.791466568156631,
+            "covered_accuracy": 0.8618807724601175,
+            "regressions": 162
+          },
+          {
+            "threshold": 0.640905,
+            "coverage": 0.8699667528629479,
+            "combined_accuracy": 0.7885112670853344,
+            "covered_accuracy": 0.8651804670912951,
+            "regressions": 154
+          },
+          {
+            "threshold": 0.655575,
+            "coverage": 0.8599926117473218,
+            "combined_accuracy": 0.7859253786479498,
+            "covered_accuracy": 0.8679123711340206,
+            "regressions": 144
+          },
+          {
+            "threshold": 0.672725,
+            "coverage": 0.8500184706316956,
+            "combined_accuracy": 0.7837089028444773,
+            "covered_accuracy": 0.8720121686223381,
+            "regressions": 137
+          },
+          {
+            "threshold": 0.686029,
+            "coverage": 0.8398596231991134,
+            "combined_accuracy": 0.7805688954562247,
+            "covered_accuracy": 0.8744226962832637,
+            "regressions": 129
+          },
+          {
+            "threshold": 0.700345,
+            "coverage": 0.8298854820834872,
+            "combined_accuracy": 0.7768747691171038,
+            "covered_accuracy": 0.8773647896728244,
+            "regressions": 124
+          },
+          {
+            "threshold": 0.712894,
+            "coverage": 0.8199113409678611,
+            "combined_accuracy": 0.773180642777983,
+            "covered_accuracy": 0.8806037395809867,
+            "regressions": 117
+          },
+          {
+            "threshold": 0.72452,
+            "coverage": 0.809937199852235,
+            "combined_accuracy": 0.7698559290727743,
+            "covered_accuracy": 0.883466362599772,
+            "regressions": 110
+          },
+          {
+            "threshold": 0.73972,
+            "coverage": 0.7999630587366088,
+            "combined_accuracy": 0.7656076837827853,
+            "covered_accuracy": 0.8857076887554837,
+            "regressions": 106
+          },
+          {
+            "threshold": 0.754053,
+            "coverage": 0.7899889176209827,
+            "combined_accuracy": 0.7615441448097525,
+            "covered_accuracy": 0.8891746551321019,
+            "regressions": 102
+          },
+          {
+            "threshold": 0.765463,
+            "coverage": 0.7800147765053564,
+            "combined_accuracy": 0.7578500184706317,
+            "covered_accuracy": 0.8924934880416765,
+            "regressions": 97
+          },
+          {
+            "threshold": 0.77653,
+            "coverage": 0.7700406353897303,
+            "combined_accuracy": 0.7524935352789065,
+            "covered_accuracy": 0.893739505876709,
+            "regressions": 93
+          },
+          {
+            "threshold": 0.785172,
+            "coverage": 0.7598817879571481,
+            "combined_accuracy": 0.7475064647210935,
+            "covered_accuracy": 0.8964511424404472,
+            "regressions": 90
+          },
+          {
+            "threshold": 0.794848,
+            "coverage": 0.749907646841522,
+            "combined_accuracy": 0.7421499815293683,
+            "covered_accuracy": 0.8972906403940887,
+            "regressions": 86
+          },
+          {
+            "threshold": 0.804995,
+            "coverage": 0.7399335057258958,
+            "combined_accuracy": 0.7380864425563354,
+            "covered_accuracy": 0.8996505242136795,
+            "regressions": 83
+          },
+          {
+            "threshold": 0.817838,
+            "coverage": 0.7299593646102697,
+            "combined_accuracy": 0.7340229035833026,
+            "covered_accuracy": 0.9015688259109311,
+            "regressions": 81
+          },
+          {
+            "threshold": 0.826393,
+            "coverage": 0.7199852234946436,
+            "combined_accuracy": 0.7286664203915774,
+            "covered_accuracy": 0.9030271934325295,
+            "regressions": 80
+          },
+          {
+            "threshold": 0.835651,
+            "coverage": 0.7100110823790173,
+            "combined_accuracy": 0.7227558182489842,
+            "covered_accuracy": 0.9037460978147763,
+            "regressions": 79
+          },
+          {
+            "threshold": 0.8443,
+            "coverage": 0.7000369412633912,
+            "combined_accuracy": 0.7183228666420391,
+            "covered_accuracy": 0.9055408970976253,
+            "regressions": 75
+          },
+          {
+            "threshold": 0.853646,
+            "coverage": 0.689878093830809,
+            "combined_accuracy": 0.7137052087181381,
+            "covered_accuracy": 0.9081659973226238,
+            "regressions": 73
+          },
+          {
+            "threshold": 0.861325,
+            "coverage": 0.6799039527151829,
+            "combined_accuracy": 0.708533431843369,
+            "covered_accuracy": 0.9092637870143983,
+            "regressions": 67
+          },
+          {
+            "threshold": 0.868863,
+            "coverage": 0.6699298115995567,
+            "combined_accuracy": 0.7026228297007757,
+            "covered_accuracy": 0.9098428453267163,
+            "regressions": 65
+          },
+          {
+            "threshold": 0.877202,
+            "coverage": 0.6599556704839306,
+            "combined_accuracy": 0.6981898780938308,
+            "covered_accuracy": 0.9129582983487265,
+            "regressions": 60
+          },
+          {
+            "threshold": 0.883933,
+            "coverage": 0.6499815293683044,
+            "combined_accuracy": 0.6928333949021056,
+            "covered_accuracy": 0.9147485080988917,
+            "regressions": 57
+          },
+          {
+            "threshold": 0.890918,
+            "coverage": 0.6400073882526782,
+            "combined_accuracy": 0.6878463243442926,
+            "covered_accuracy": 0.9163059163059163,
+            "regressions": 56
+          },
+          {
+            "threshold": 0.895454,
+            "coverage": 0.6300332471370521,
+            "combined_accuracy": 0.6839674916882157,
+            "covered_accuracy": 0.9187921430665494,
+            "regressions": 52
+          },
+          {
+            "threshold": 0.900932,
+            "coverage": 0.6200591060214259,
+            "combined_accuracy": 0.6782415958625785,
+            "covered_accuracy": 0.9192731605600238,
+            "regressions": 51
+          },
+          {
+            "threshold": 0.905733,
+            "coverage": 0.6099002585888438,
+            "combined_accuracy": 0.6738086442556336,
+            "covered_accuracy": 0.922168382798304,
+            "regressions": 50
+          },
+          {
+            "threshold": 0.911608,
+            "coverage": 0.5999261174732176,
+            "combined_accuracy": 0.6695603989656447,
+            "covered_accuracy": 0.9245689655172413,
+            "regressions": 48
+          },
+          {
+            "threshold": 0.916809,
+            "coverage": 0.5899519763575914,
+            "combined_accuracy": 0.6638345031400074,
+            "covered_accuracy": 0.926111458985598,
+            "regressions": 45
+          },
+          {
+            "threshold": 0.920427,
+            "coverage": 0.5799778352419652,
+            "combined_accuracy": 0.657739194680458,
+            "covered_accuracy": 0.9273885350318471,
+            "regressions": 42
+          },
+          {
+            "threshold": 0.925197,
+            "coverage": 0.5700036941263391,
+            "combined_accuracy": 0.6536756557074251,
+            "covered_accuracy": 0.9309786130913804,
+            "regressions": 36
+          },
+          {
+            "threshold": 0.929965,
+            "coverage": 0.560029553010713,
+            "combined_accuracy": 0.648503878832656,
+            "covered_accuracy": 0.933377308707124,
+            "regressions": 34
+          },
+          {
+            "threshold": 0.933174,
+            "coverage": 0.5500554118950868,
+            "combined_accuracy": 0.6433321019578869,
+            "covered_accuracy": 0.9358629952988583,
+            "regressions": 31
+          },
+          {
+            "threshold": 0.93709,
+            "coverage": 0.5400812707794607,
+            "combined_accuracy": 0.6377909124492057,
+            "covered_accuracy": 0.9374145006839946,
+            "regressions": 30
+          },
+          {
+            "threshold": 0.940724,
+            "coverage": 0.5299224233468784,
+            "combined_accuracy": 0.6328038418913927,
+            "covered_accuracy": 0.9393516904844894,
+            "regressions": 26
+          },
+          {
+            "threshold": 0.944268,
+            "coverage": 0.5199482822312523,
+            "combined_accuracy": 0.6268932397487994,
+            "covered_accuracy": 0.9399644760213144,
+            "regressions": 25
+          },
+          {
+            "threshold": 0.948099,
+            "coverage": 0.5099741411156261,
+            "combined_accuracy": 0.6215367565570743,
+            "covered_accuracy": 0.9406012314378849,
+            "regressions": 24
+          },
+          {
+            "threshold": 0.951436,
+            "coverage": 0.5,
+            "combined_accuracy": 0.6161802733653491,
+            "covered_accuracy": 0.9408939785740672,
+            "regressions": 24
+          },
+          {
+            "threshold": 0.954023,
+            "coverage": 0.49002585888437383,
+            "combined_accuracy": 0.6126708533431844,
+            "covered_accuracy": 0.944214097248398,
+            "regressions": 20
+          },
+          {
+            "threshold": 0.956498,
+            "coverage": 0.48005171776874767,
+            "combined_accuracy": 0.6078684891023273,
+            "covered_accuracy": 0.9453636013851482,
+            "regressions": 18
+          },
+          {
+            "threshold": 0.958537,
+            "coverage": 0.47007757665312155,
+            "combined_accuracy": 0.6026967122275582,
+            "covered_accuracy": 0.9469548133595285,
+            "regressions": 15
+          },
+          {
+            "threshold": 0.960376,
+            "coverage": 0.45991872922053934,
+            "combined_accuracy": 0.597709641669745,
+            "covered_accuracy": 0.948995983935743,
+            "regressions": 13
+          },
+          {
+            "threshold": 0.963265,
+            "coverage": 0.44994458810491317,
+            "combined_accuracy": 0.5919837458441078,
+            "covered_accuracy": 0.9499178981937603,
+            "regressions": 12
+          },
+          {
+            "threshold": 0.965379,
+            "coverage": 0.43997044698928706,
+            "combined_accuracy": 0.5860731437015146,
+            "covered_accuracy": 0.9500419815281276,
+            "regressions": 12
+          },
+          {
+            "threshold": 0.96692,
+            "coverage": 0.4299963058736609,
+            "combined_accuracy": 0.5799778352419652,
+            "covered_accuracy": 0.9510309278350515,
+            "regressions": 12
+          },
+          {
+            "threshold": 0.968873,
+            "coverage": 0.4200221647580347,
+            "combined_accuracy": 0.5744366457332841,
+            "covered_accuracy": 0.951627088830255,
+            "regressions": 9
+          },
+          {
+            "threshold": 0.970763,
+            "coverage": 0.41004802364240855,
+            "combined_accuracy": 0.5685260435906908,
+            "covered_accuracy": 0.9518018018018019,
+            "regressions": 9
+          },
+          {
+            "threshold": 0.972469,
+            "coverage": 0.40007388252678244,
+            "combined_accuracy": 0.5626154414480975,
+            "covered_accuracy": 0.9519852262234534,
+            "regressions": 9
+          },
+          {
+            "threshold": 0.974047,
+            "coverage": 0.3900997414111563,
+            "combined_accuracy": 0.5567048393055043,
+            "covered_accuracy": 0.9535984848484849,
+            "regressions": 7
+          },
+          {
+            "threshold": 0.975516,
+            "coverage": 0.37994089397857406,
+            "combined_accuracy": 0.5520871813816033,
+            "covered_accuracy": 0.9533300923675255,
+            "regressions": 7
+          },
+          {
+            "threshold": 0.976913,
+            "coverage": 0.3699667528629479,
+            "combined_accuracy": 0.5469154045068342,
+            "covered_accuracy": 0.9550673989016475,
+            "regressions": 6
+          },
+          {
+            "threshold": 0.978492,
+            "coverage": 0.3599926117473218,
+            "combined_accuracy": 0.5421130402659771,
+            "covered_accuracy": 0.9563878912262699,
+            "regressions": 6
+          },
+          {
+            "threshold": 0.979541,
+            "coverage": 0.3500184706316956,
+            "combined_accuracy": 0.5362024381233839,
+            "covered_accuracy": 0.9551451187335093,
+            "regressions": 6
+          },
+          {
+            "threshold": 0.980824,
+            "coverage": 0.34004432951606944,
+            "combined_accuracy": 0.5308459549316586,
+            "covered_accuracy": 0.955458989679522,
+            "regressions": 6
+          },
+          {
+            "threshold": 0.981792,
+            "coverage": 0.3300701884004433,
+            "combined_accuracy": 0.5247506464721093,
+            "covered_accuracy": 0.9557918298824846,
+            "regressions": 6
+          },
+          {
+            "threshold": 0.98304,
+            "coverage": 0.32009604728481716,
+            "combined_accuracy": 0.5205024011821204,
+            "covered_accuracy": 0.9561454125793422,
+            "regressions": 4
+          },
+          {
+            "threshold": 0.984012,
+            "coverage": 0.30993719985223495,
+            "combined_accuracy": 0.5157000369412634,
+            "covered_accuracy": 0.9576877234803337,
+            "regressions": 4
+          },
+          {
+            "threshold": 0.984861,
+            "coverage": 0.2999630587366088,
+            "combined_accuracy": 0.5099741411156261,
+            "covered_accuracy": 0.9568965517241379,
+            "regressions": 4
+          },
+          {
+            "threshold": 0.985914,
+            "coverage": 0.2899889176209826,
+            "combined_accuracy": 0.5060953084595493,
+            "covered_accuracy": 0.9573248407643312,
+            "regressions": 4
+          },
+          {
+            "threshold": 0.986784,
+            "coverage": 0.2800147765053565,
+            "combined_accuracy": 0.5007388252678242,
+            "covered_accuracy": 0.9597625329815304,
+            "regressions": 4
+          },
+          {
+            "threshold": 0.98768,
+            "coverage": 0.27004063538973033,
+            "combined_accuracy": 0.49556704839305504,
+            "covered_accuracy": 0.9596443228454172,
+            "regressions": 4
+          },
+          {
+            "threshold": 0.988424,
+            "coverage": 0.26006649427410417,
+            "combined_accuracy": 0.4915035094200222,
+            "covered_accuracy": 0.9630681818181818,
+            "regressions": 3
+          },
+          {
+            "threshold": 0.989225,
+            "coverage": 0.250092353158478,
+            "combined_accuracy": 0.48762467676394533,
+            "covered_accuracy": 0.96602658788774,
+            "regressions": 2
+          },
+          {
+            "threshold": 0.989847,
+            "coverage": 0.24011821204285186,
+            "combined_accuracy": 0.4826376062061323,
+            "covered_accuracy": 0.9661538461538461,
+            "regressions": 2
+          },
+          {
+            "threshold": 0.990648,
+            "coverage": 0.22995936461026967,
+            "combined_accuracy": 0.4778352419652752,
+            "covered_accuracy": 0.9654618473895582,
+            "regressions": 2
+          },
+          {
+            "threshold": 0.991326,
+            "coverage": 0.21998522349464353,
+            "combined_accuracy": 0.47340229035833026,
+            "covered_accuracy": 0.964735516372796,
+            "regressions": 2
+          },
+          {
+            "threshold": 0.991868,
+            "coverage": 0.21001108237901736,
+            "combined_accuracy": 0.4684152198005172,
+            "covered_accuracy": 0.9665787159190853,
+            "regressions": 2
+          },
+          {
+            "threshold": 0.992456,
+            "coverage": 0.20003694126339122,
+            "combined_accuracy": 0.46490579977835245,
+            "covered_accuracy": 0.9676823638042474,
+            "regressions": 1
+          },
+          {
+            "threshold": 0.993025,
+            "coverage": 0.19006280014776505,
+            "combined_accuracy": 0.46121167343923164,
+            "covered_accuracy": 0.966958211856171,
+            "regressions": 1
+          },
+          {
+            "threshold": 0.993511,
+            "coverage": 0.1800886590321389,
+            "combined_accuracy": 0.4569634281492427,
+            "covered_accuracy": 0.9661538461538461,
+            "regressions": 1
+          },
+          {
+            "threshold": 0.993984,
+            "coverage": 0.17011451791651275,
+            "combined_accuracy": 0.4528998891762098,
+            "covered_accuracy": 0.9674267100977199,
+            "regressions": 1
+          },
+          {
+            "threshold": 0.99437,
+            "coverage": 0.15995567048393056,
+            "combined_accuracy": 0.44883635020317697,
+            "covered_accuracy": 0.9665127020785219,
+            "regressions": 1
+          },
+          {
+            "threshold": 0.994859,
+            "coverage": 0.15016623568526044,
+            "combined_accuracy": 0.4449575175471001,
+            "covered_accuracy": 0.9692496924969249,
+            "regressions": 1
+          },
+          {
+            "threshold": 0.995294,
+            "coverage": 0.14000738825267825,
+            "combined_accuracy": 0.44200221647580346,
+            "covered_accuracy": 0.9709762532981531,
+            "regressions": 0
+          },
+          {
+            "threshold": 0.995697,
+            "coverage": 0.13021795345400813,
+            "combined_accuracy": 0.43867750277059475,
+            "covered_accuracy": 0.9730496453900709,
+            "regressions": 0
+          },
+          {
+            "threshold": 0.996291,
+            "coverage": 0.12005910602142593,
+            "combined_accuracy": 0.43590690801625415,
+            "covered_accuracy": 0.9738461538461538,
+            "regressions": 0
+          },
+          {
+            "threshold": 0.996686,
+            "coverage": 0.11008496490579978,
+            "combined_accuracy": 0.43202807536017734,
+            "covered_accuracy": 0.9714765100671141,
+            "regressions": 0
+          },
+          {
+            "threshold": 0.99705,
+            "coverage": 0.10011082379017362,
+            "combined_accuracy": 0.4287033616549686,
+            "covered_accuracy": 0.974169741697417,
+            "regressions": 0
+          },
+          {
+            "threshold": 0.997386,
+            "coverage": 0.09013668267454747,
+            "combined_accuracy": 0.425748060583672,
+            "covered_accuracy": 0.9774590163934426,
+            "regressions": 0
+          },
+          {
+            "threshold": 0.997703,
+            "coverage": 0.08016254155892132,
+            "combined_accuracy": 0.4239009974141116,
+            "covered_accuracy": 0.9746543778801844,
+            "regressions": 0
+          },
+          {
+            "threshold": 0.997964,
+            "coverage": 0.07000369412633912,
+            "combined_accuracy": 0.4218692279275951,
+            "covered_accuracy": 0.978891820580475,
+            "regressions": 0
+          },
+          {
+            "threshold": 0.998248,
+            "coverage": 0.060029553010712965,
+            "combined_accuracy": 0.4200221647580347,
+            "covered_accuracy": 0.9907692307692307,
+            "regressions": 0
+          },
+          {
+            "threshold": 0.998595,
+            "coverage": 0.05005541189508681,
+            "combined_accuracy": 0.41688215736978207,
+            "covered_accuracy": 0.988929889298893,
+            "regressions": 0
+          },
+          {
+            "threshold": 0.998848,
+            "coverage": 0.04008127077946066,
+            "combined_accuracy": 0.41466568156630956,
+            "covered_accuracy": 0.9953917050691244,
+            "regressions": 0
+          },
+          {
+            "threshold": 0.99905,
+            "coverage": 0.029922423346878464,
+            "combined_accuracy": 0.4130033247137052,
+            "covered_accuracy": 0.9938271604938271,
+            "regressions": 0
+          },
+          {
+            "threshold": 0.99924,
+            "coverage": 0.020132988548208348,
+            "combined_accuracy": 0.4107868489102327,
+            "covered_accuracy": 0.9908256880733946,
+            "regressions": 0
+          },
+          {
+            "threshold": 0.999454,
+            "coverage": 0.009974141115626155,
+            "combined_accuracy": 0.40801625415589216,
+            "covered_accuracy": 1.0,
+            "regressions": 0
+          },
+          {
+            "threshold": 0.99978,
+            "coverage": 0.0,
+            "combined_accuracy": 0.4059844846693757,
+            "covered_accuracy": null,
+            "regressions": 0
+          }
+        ],
+        "baseline_accuracy": 0.4059844846693757,
+        "no_abstention_accuracy": 0.8134466198743997,
+        "best_combined": {
+          "threshold": 0.287819,
+          "coverage": 0.999815293683044,
+          "combined_accuracy": 0.8134466198743997,
+          "covered_accuracy": 0.8135968963606134,
+          "regressions": 256
+        },
+        "conservative_zero_regression": {
+          "threshold": 0.995294,
+          "coverage": 0.14000738825267825,
+          "combined_accuracy": 0.44200221647580346,
+          "covered_accuracy": 0.9709762532981531,
+          "regressions": 0
+        },
+        "ece": 0.0576062929816676
+      },
+      "nbest_coverage": {
+        "all": {
+          "1": 0.8134466198743997,
+          "2": 0.9436645733284078,
+          "3": 0.9804211304026598,
+          "4": 0.9918729220539343,
+          "5": 0.9970446989287034
+        },
+        "variation_only": {
+          "1": 0.6122448979591837,
+          "2": 0.9387755102040817,
+          "3": 0.9795918367346939,
+          "4": 1.0,
+          "5": 1.0
+        },
+        "n_variation": 49
+      },
+      "operating_point_dev_chosen": {
+        "chosen_on": "dev split, same condition",
+        "max_accuracy": {
+          "threshold": 0.242734,
+          "coverage": 1.0,
+          "combined_accuracy": 0.8134466198743997,
+          "regressions": 256,
+          "improvements": 2462,
+          "n": 5414
+        },
+        "zero_regression_on_dev": {
+          "threshold": 0.997893,
+          "coverage": 0.07258958256372368,
+          "combined_accuracy": 0.42223864056150723,
+          "regressions": 0,
+          "improvements": 88,
+          "n": 5414
+        },
+        "dev_zero_regression_coverage": 0.08007448789571694
+      }
+    }
+  },
+  "notes": {
+    "fairness": "Accuracy is computed on items where OpenJTalk's own reading matched the gold reading, so the comparison is about accent rather than G2P.",
+    "downstep_condition_invariance": "Downstep's prediction does not depend on the condition: it reads components, never the compound's own dictionary entry.",
+    "no_human_annotation": "All gold is dictionary-derived. No native-speaker annotation was collected for this release.",
+    "comparable_set": "Strata are scored on the 5,414 test compounds whose reading OpenJTalk produced correctly in BOTH the listed and the unlisted condition, so the two columns describe the same words."
+  }
+}

mora_validation.json

@@ -0,0 +1,1083 @@
+{
+  "tool": "scripts/validate_mora.py",
+  "generated_at": "2026-09-03T16:16:51+0900",
+  "elapsed_seconds": 19.26,
+  "environment": {
+    "python": "3.13.5",
+    "platform": "macOS-26.6.2-arm64-arm-64bit-Mach-O"
+  },
+  "tokenizer": {
+    "module": "downstep.mora",
+    "fold_small_counters": true,
+    "strict": false,
+    "rules": {
+      "small_kana_fuse_with_the_preceding_kana": [
+        "ァ",
+        "ィ",
+        "ゥ",
+        "ェ",
+        "ォ",
+        "ャ",
+        "ュ",
+        "ョ",
+        "ヮ"
+      ],
+      "own_mora": [
+        "ッ",
+        "ン",
+        "ー"
+      ],
+      "small_counters_folded_to_full": {
+        "ヵ": "カ",
+        "ヶ": "ケ"
+      }
+    }
+  },
+  "documented_examples": {
+    "note": "definitions taken from downstep/mora.py's own docstring plus mechanical applications of the four rules it states; not a measurement",
+    "n": 29,
+    "passed": 29,
+    "failed": 0,
+    "failures": []
+  },
+  "losslessness": {
+    "checked": 788331,
+    "lossless": 788331,
+    "n_failures": 0,
+    "failures_shown": []
+  },
+  "naist": {
+    "source": "NAIST Japanese Dictionary (naist-jdic), as shipped with Open JTalk",
+    "check": "direct: tokenizer mora count vs the dictionary's own accmora denominator",
+    "devoicing_marks_stripped": [
+      "’",
+      "'"
+    ],
+    "entries_read": 788331,
+    "entries_without_recorded_count": 0,
+    "entries_with_recorded_count": 788331,
+    "agree": 788190,
+    "disagree": 141,
+    "agreement_rate": 0.9998211411196566,
+    "delta_histogram": {
+      "1": 138,
+      "2": 1,
+      "3": 2
+    },
+    "attribution": {
+      "rule": "a kana pair the tokenizer fuses into one mora is blamed only if it occurs in zero agreeing entries; the dictionary then evidently counts it as two. Completeness is verified by recounting every disagreeing entry under that assumption.",
+      "explained": 141,
+      "unexplained": 0,
+      "unexplained_rows": [],
+      "n_pairs": 33,
+      "pairs": [
+        {
+          "mora": "フュ",
+          "mechanism": "palatalisation_on_a_non_standard_base",
+          "disagreeing_rows_containing_it": 67,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "ふゅーじょん",
+              "reading": "フュージョン",
+              "dict_moras": 5,
+              "tokenizer_moras": 4
+            },
+            {
+              "surface": "アフュス",
+              "reading": "アフュス",
+              "dict_moras": 4,
+              "tokenizer_moras": 3
+            },
+            {
+              "surface": "アフュティス",
+              "reading": "アフュティス",
+              "dict_moras": 5,
+              "tokenizer_moras": 4
+            },
+            {
+              "surface": "インフューザー",
+              "reading": "インフューザー",
+              "dict_moras": 7,
+              "tokenizer_moras": 6
+            }
+          ]
+        },
+        {
+          "mora": "イュ",
+          "mechanism": "palatalisation_on_a_non_standard_base",
+          "disagreeing_rows_containing_it": 12,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "かいい",
+              "reading": "カイュー",
+              "dict_moras": 4,
+              "tokenizer_moras": 3
+            },
+            {
+              "surface": "かっわいい",
+              "reading": "カッワイュー",
+              "dict_moras": 6,
+              "tokenizer_moras": 5
+            },
+            {
+              "surface": "かわいい",
+              "reading": "カワイュウ",
+              "dict_moras": 5,
+              "tokenizer_moras": 4
+            },
+            {
+              "surface": "かわいい",
+              "reading": "カワイュウ",
+              "dict_moras": 5,
+              "tokenizer_moras": 4
+            }
+          ]
+        },
+        {
+          "mora": "クォ",
+          "mechanism": "small_vowel_after_a_full_kana",
+          "disagreeing_rows_containing_it": 5,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "クォータリー",
+              "reading": "クォータリー",
+              "dict_moras": 6,
+              "tokenizer_moras": 5
+            },
+            {
+              "surface": "クォーター",
+              "reading": "クォーター",
+              "dict_moras": 5,
+              "tokenizer_moras": 4
+            },
+            {
+              "surface": "クォーテーション",
+              "reading": "クォーテーション",
+              "dict_moras": 7,
+              "tokenizer_moras": 6
+            },
+            {
+              "surface": "クォート",
+              "reading": "クォート",
+              "dict_moras": 4,
+              "tokenizer_moras": 3
+            }
+          ]
+        },
+        {
+          "mora": "シィ",
+          "mechanism": "small_vowel_after_a_full_kana",
+          "disagreeing_rows_containing_it": 5,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "アイディシィソリューションズ",
+              "reading": "アイディシィソリューションズ",
+              "dict_moras": 11,
+              "tokenizer_moras": 10
+            },
+            {
+              "surface": "エフ・ディ・シィ・プロダクツ",
+              "reading": "エフディシィプロダクツ",
+              "dict_moras": 10,
+              "tokenizer_moras": 9
+            },
+            {
+              "surface": "ルートフィッシィング",
+              "reading": "ルートフィッシィング",
+              "dict_moras": 9,
+              "tokenizer_moras": 8
+            },
+            {
+              "surface": "レガシィ",
+              "reading": "レガシィ",
+              "dict_moras": 4,
+              "tokenizer_moras": 3
+            }
+          ]
+        },
+        {
+          "mora": "ヂャ",
+          "mechanism": "youon_written_on_the_di_row_historical_orthography",
+          "disagreeing_rows_containing_it": 5,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "ぢゃぢゅぢょ",
+              "reading": "ヂャヂュヂョ",
+              "dict_moras": 6,
+              "tokenizer_moras": 3
+            },
+            {
+              "surface": "ヂャヂュヂョ",
+              "reading": "ヂャヂュヂョ",
+              "dict_moras": 6,
+              "tokenizer_moras": 3
+            },
+            {
+              "surface": "茶",
+              "reading": "ヂャ",
+              "dict_moras": 2,
+              "tokenizer_moras": 1
+            },
+            {
+              "surface": "茶屋",
+              "reading": "ヂャヤ",
+              "dict_moras": 3,
+              "tokenizer_moras": 2
+            }
+          ]
+        },
+        {
+          "mora": "ヂョ",
+          "mechanism": "youon_written_on_the_di_row_historical_orthography",
+          "disagreeing_rows_containing_it": 5,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "ぢゃぢゅぢょ",
+              "reading": "ヂャヂュヂョ",
+              "dict_moras": 6,
+              "tokenizer_moras": 3
+            },
+            {
+              "surface": "ヂャヂュヂョ",
+              "reading": "ヂャヂュヂョ",
+              "dict_moras": 6,
+              "tokenizer_moras": 3
+            },
+            {
+              "surface": "提灯",
+              "reading": "ヂョウチン",
+              "dict_moras": 5,
+              "tokenizer_moras": 4
+            },
+            {
+              "surface": "提灯",
+              "reading": "ヂョウチン",
+              "dict_moras": 5,
+              "tokenizer_moras": 4
+            }
+          ]
+        },
+        {
+          "mora": "クィ",
+          "mechanism": "small_vowel_after_a_full_kana",
+          "disagreeing_rows_containing_it": 4,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "クィック",
+              "reading": "クィック",
+              "dict_moras": 4,
+              "tokenizer_moras": 3
+            },
+            {
+              "surface": "クィックソート",
+              "reading": "クィックソート",
+              "dict_moras": 7,
+              "tokenizer_moras": 6
+            },
+            {
+              "surface": "クィンテット",
+              "reading": "クィンテット",
+              "dict_moras": 6,
+              "tokenizer_moras": 5
+            },
+            {
+              "surface": "ラブリークィーン",
+              "reading": "ラブリークィーン",
+              "dict_moras": 8,
+              "tokenizer_moras": 7
+            }
+          ]
+        },
+        {
+          "mora": "ナュ",
+          "mechanism": "palatalisation_on_a_non_standard_base",
+          "disagreeing_rows_containing_it": 4,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "とんでもない",
+              "reading": "トンデモナュウ",
+              "dict_moras": 7,
+              "tokenizer_moras": 6
+            },
+            {
+              "surface": "とんでもない",
+              "reading": "トンデモナュウ",
+              "dict_moras": 7,
+              "tokenizer_moras": 6
+            },
+            {
+              "surface": "無い",
+              "reading": "ナュー",
+              "dict_moras": 3,
+              "tokenizer_moras": 2
+            },
+            {
+              "surface": "無い",
+              "reading": "ナュー",
+              "dict_moras": 3,
+              "tokenizer_moras": 2
+            }
+          ]
+        },
+        {
+          "mora": "クァ",
+          "mechanism": "small_vowel_after_a_full_kana",
+          "disagreeing_rows_containing_it": 3,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "クァルテット",
+              "reading": "クァルテット",
+              "dict_moras": 6,
+              "tokenizer_moras": 5
+            },
+            {
+              "surface": "クァンジュ",
+              "reading": "クァンジュ",
+              "dict_moras": 4,
+              "tokenizer_moras": 3
+            },
+            {
+              "surface": "高広鎮",
+              "reading": "コークァンジン",
+              "dict_moras": 7,
+              "tokenizer_moras": 6
+            }
+          ]
+        },
+        {
+          "mora": "クェ",
+          "mechanism": "small_vowel_after_a_full_kana",
+          "disagreeing_rows_containing_it": 3,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "クェスチョン",
+              "reading": "クェスチョン",
+              "dict_moras": 5,
+              "tokenizer_moras": 4
+            },
+            {
+              "surface": "クェッション",
+              "reading": "クェッション",
+              "dict_moras": 5,
+              "tokenizer_moras": 4
+            },
+            {
+              "surface": "ＡＦＬＡＣスクェア",
+              "reading": "アフラックスクェア",
+              "dict_moras": 9,
+              "tokenizer_moras": 8
+            }
+          ]
+        },
+        {
+          "mora": "グァ",
+          "mechanism": "small_vowel_after_a_full_kana",
+          "disagreeing_rows_containing_it": 3,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "グァラニ",
+              "reading": "グァラニ",
+              "dict_moras": 4,
+              "tokenizer_moras": 3
+            },
+            {
+              "surface": "グァラニ",
+              "reading": "グァラニ",
+              "dict_moras": 4,
+              "tokenizer_moras": 3
+            },
+            {
+              "surface": "パラグアイグァラニ",
+              "reading": "パラグアイグァラニ",
+              "dict_moras": 9,
+              "tokenizer_moras": 8
+            }
+          ]
+        },
+        {
+          "mora": "ジァ",
+          "mechanism": "small_vowel_after_a_full_kana",
+          "disagreeing_rows_containing_it": 3,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "ジァン",
+              "reading": "ジァン",
+              "dict_moras": 3,
+              "tokenizer_moras": 2
+            },
+            {
+              "surface": "ジァン・ジァン",
+              "reading": "ジァンジァン",
+              "dict_moras": 6,
+              "tokenizer_moras": 4
+            }
+          ]
+        },
+        {
+          "mora": "ビィ",
+          "mechanism": "small_vowel_after_a_full_kana",
+          "disagreeing_rows_containing_it": 3,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "カービィ",
+              "reading": "カービィ",
+              "dict_moras": 4,
+              "tokenizer_moras": 3
+            },
+            {
+              "surface": "ギャツビィゴルフクラブ",
+              "reading": "ギャツビィゴルフクラブ",
+              "dict_moras": 10,
+              "tokenizer_moras": 9
+            },
+            {
+              "surface": "シューメーカー・レビィ",
+              "reading": "シューメーカーレビィ",
+              "dict_moras": 9,
+              "tokenizer_moras": 8
+            }
+          ]
+        },
+        {
+          "mora": "ケュ",
+          "mechanism": "palatalisation_on_a_non_standard_base",
+          "disagreeing_rows_containing_it": 2,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "寒けい",
+              "reading": "サムケュー",
+              "dict_moras": 5,
+              "tokenizer_moras": 4
+            },
+            {
+              "surface": "鮮らけい",
+              "reading": "アザラケュー",
+              "dict_moras": 6,
+              "tokenizer_moras": 5
+            }
+          ]
+        },
+        {
+          "mora": "ジィ",
+          "mechanism": "small_vowel_after_a_full_kana",
+          "disagreeing_rows_containing_it": 2,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "パンジィ",
+              "reading": "パンジィ",
+              "dict_moras": 4,
+              "tokenizer_moras": 3
+            },
+            {
+              "surface": "ファジィ",
+              "reading": "ファジィ",
+              "dict_moras": 3,
+              "tokenizer_moras": 2
+            }
+          ]
+        },
+        {
+          "mora": "ヂュ",
+          "mechanism": "youon_written_on_the_di_row_historical_orthography",
+          "disagreeing_rows_containing_it": 2,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "ぢゃぢゅぢょ",
+              "reading": "ヂャヂュヂョ",
+              "dict_moras": 6,
+              "tokenizer_moras": 3
+            },
+            {
+              "surface": "ヂャヂュヂョ",
+              "reading": "ヂャヂュヂョ",
+              "dict_moras": 6,
+              "tokenizer_moras": 3
+            }
+          ]
+        },
+        {
+          "mora": "デェ",
+          "mechanism": "small_vowel_after_a_full_kana",
+          "disagreeing_rows_containing_it": 2,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "でぇ",
+              "reading": "デェ",
+              "dict_moras": 2,
+              "tokenizer_moras": 1
+            },
+            {
+              "surface": "デェ",
+              "reading": "デェ",
+              "dict_moras": 2,
+              "tokenizer_moras": 1
+            }
+          ]
+        },
+        {
+          "mora": "エィ",
+          "mechanism": "small_vowel_after_a_full_kana",
+          "disagreeing_rows_containing_it": 1,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "エィンデン",
+              "reading": "エィンデン",
+              "dict_moras": 5,
+              "tokenizer_moras": 4
+            }
+          ]
+        },
+        {
+          "mora": "ガュ",
+          "mechanism": "palatalisation_on_a_non_standard_base",
+          "disagreeing_rows_containing_it": 1,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "凶い",
+              "reading": "マガュー",
+              "dict_moras": 4,
+              "tokenizer_moras": 3
+            }
+          ]
+        },
+        {
+          "mora": "キィ",
+          "mechanism": "small_vowel_after_a_full_kana",
+          "disagreeing_rows_containing_it": 1,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "ラッキィ池田",
+              "reading": "ラッキィイケダ",
+              "dict_moras": 7,
+              "tokenizer_moras": 6
+            }
+          ]
+        },
+        {
+          "mora": "グゥ",
+          "mechanism": "small_vowel_after_a_full_kana",
+          "disagreeing_rows_containing_it": 1,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "アウグゥストゥス",
+              "reading": "アウグゥストゥス",
+              "dict_moras": 7,
+              "tokenizer_moras": 6
+            }
+          ]
+        },
+        {
+          "mora": "ゲュ",
+          "mechanism": "palatalisation_on_a_non_standard_base",
+          "disagreeing_rows_containing_it": 1,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "繁い",
+              "reading": "シゲュー",
+              "dict_moras": 4,
+              "tokenizer_moras": 3
+            }
+          ]
+        },
+        {
+          "mora": "コァ",
+          "mechanism": "small_vowel_after_a_full_kana",
+          "disagreeing_rows_containing_it": 1,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "黄家駒",
+              "reading": "コァンジャジュ",
+              "dict_moras": 5,
+              "tokenizer_moras": 4
+            }
+          ]
+        },
+        {
+          "mora": "ゴォ",
+          "mechanism": "small_vowel_after_a_full_kana",
+          "disagreeing_rows_containing_it": 1,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "ゴォォール",
+              "reading": "ゴォォール",
+              "dict_moras": 5,
+              "tokenizer_moras": 4
+            }
+          ]
+        },
+        {
+          "mora": "スェ",
+          "mechanism": "small_vowel_after_a_full_kana",
+          "disagreeing_rows_containing_it": 1,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "スェーター",
+              "reading": "スェーター",
+              "dict_moras": 5,
+              "tokenizer_moras": 4
+            }
+          ]
+        },
+        {
+          "mora": "セュ",
+          "mechanism": "palatalisation_on_a_non_standard_base",
+          "disagreeing_rows_containing_it": 1,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "狭い",
+              "reading": "セュー",
+              "dict_moras": 3,
+              "tokenizer_moras": 2
+            }
+          ]
+        },
+        {
+          "mora": "チァ",
+          "mechanism": "small_vowel_after_a_full_kana",
+          "disagreeing_rows_containing_it": 1,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "江蘇",
+              "reading": "チァンスー",
+              "dict_moras": 5,
+              "tokenizer_moras": 4
+            }
+          ]
+        },
+        {
+          "mora": "テェ",
+          "mechanism": "small_vowel_after_a_full_kana",
+          "disagreeing_rows_containing_it": 1,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "いたい",
+              "reading": "イテェ",
+              "dict_moras": 3,
+              "tokenizer_moras": 2
+            }
+          ]
+        },
+        {
+          "mora": "ネュ",
+          "mechanism": "palatalisation_on_a_non_standard_base",
+          "disagreeing_rows_containing_it": 1,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "おっかねい",
+              "reading": "オッカネュー",
+              "dict_moras": 6,
+              "tokenizer_moras": 5
+            }
+          ]
+        },
+        {
+          "mora": "フョ",
+          "mechanism": "palatalisation_on_a_non_standard_base",
+          "disagreeing_rows_containing_it": 1,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "フョードロフ",
+              "reading": "フョードロフ",
+              "dict_moras": 6,
+              "tokenizer_moras": 5
+            }
+          ]
+        },
+        {
+          "mora": "ムュ",
+          "mechanism": "palatalisation_on_a_non_standard_base",
+          "disagreeing_rows_containing_it": 1,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "薄ら眠い",
+              "reading": "ウスラネムュー",
+              "dict_moras": 7,
+              "tokenizer_moras": 6
+            }
+          ]
+        },
+        {
+          "mora": "モュ",
+          "mechanism": "palatalisation_on_a_non_standard_base",
+          "disagreeing_rows_containing_it": 1,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "パンケモュウパロ川",
+              "reading": "パンケモュウパロガワ",
+              "dict_moras": 10,
+              "tokenizer_moras": 9
+            }
+          ]
+        },
+        {
+          "mora": "ルュ",
+          "mechanism": "palatalisation_on_a_non_standard_base",
+          "disagreeing_rows_containing_it": 1,
+          "agreeing_rows_containing_it": 0,
+          "examples": [
+            {
+              "surface": "アルュメ",
+              "reading": "アルュメ",
+              "dict_moras": 4,
+              "tokenizer_moras": 3
+            }
+          ]
+        }
+      ],
+      "mechanism_groups": {
+        "palatalisation_on_a_non_standard_base": {
+          "pairs": [
+            "イュ",
+            "ガュ",
+            "ケュ",
+            "ゲュ",
+            "セュ",
+            "ナュ",
+            "ネュ",
+            "フュ",
+            "フョ",
+            "ムュ",
+            "モュ",
+            "ルュ"
+          ],
+          "mismatch_rows_touched": 93
+        },
+        "small_vowel_after_a_full_kana": {
+          "pairs": [
+            "エィ",
+            "キィ",
+            "クァ",
+            "クィ",
+            "クェ",
+            "クォ",
+            "グァ",
+            "グゥ",
+            "コァ",
+            "ゴォ",
+            "シィ",
+            "ジァ",
+            "ジィ",
+            "スェ",
+            "チァ",
+            "テェ",
+            "デェ",
+            "ビィ"
+          ],
+          "mismatch_rows_touched": 41
+        },
+        "youon_written_on_the_di_row_historical_orthography": {
+          "pairs": [
+            "ヂャ",
+            "ヂュ",
+            "ヂョ"
+          ],
+          "mismatch_rows_touched": 12
+        }
+      },
+      "surface_script_of_disagreeing_entries": {
+        "kana_only": 110,
+        "contains_kanji": 27,
+        "contains_latin": 4
+      }
+    },
+    "nucleus_within_tokenizer_mora_count": {
+      "note": "sanity check only; NAIST accmora is self-consistent by construction",
+      "n": 788331,
+      "in_range": 788331,
+      "rate": 1.0
+    }
+  },
+  "unidic": {
+    "mode": "sys.dic",
+    "source": "UniDic 2.1.2 (unidic-mecab)",
+    "check": "indirect BOUND, not an equality: UniDic records no mora count, so the test is 0 <= aType <= mora_count(reading). It detects under-counting and is blind to over-counting.",
+    "devoicing_marks_stripped": [
+      "’",
+      "'"
+    ],
+    "entries_with_accent": 751705,
+    "accent_annotations": 901116,
+    "within_bound_using_surface_reading": 900945,
+    "within_bound_rate_surface_reading": 0.9998102353082178,
+    "out_of_bound_using_surface_reading": 171,
+    "explained_by_lemma_length": {
+      "note": "UniDic states aType for the lemma; an inflected surface form is shorter than its own lemma. Re-checking against pronBase absorbs these.",
+      "count": 95,
+      "by_pos": {
+        "動詞-一般": 93,
+        "形容詞-一般": 2
+      }
+    },
+    "residual_not_attributable_to_the_tokenizer": {
+      "note": "the recorded aType exceeds the mora count of BOTH reading fields in the same row, so no segmentation of that row's own text can satisfy it; these are annotation slips in the source dictionary",
+      "count": 76,
+      "rate_of_all_annotations": 8.43398630143067e-05,
+      "by_pos": {
+        "動詞-一般": 39,
+        "名詞-普通名詞": 19,
+        "名詞-固有名詞": 11,
+        "名詞-数詞": 4,
+        "副詞-*": 3
+      },
+      "examples": [
+        {
+          "orth": "すま",
+          "pron": "スマ",
+          "pron_base": "スマ",
+          "recorded_aType": 4,
+          "tokenizer_moras_pron": 2,
+          "tokenizer_moras_lemma": 2,
+          "pos": "名詞-普通名詞"
+        },
+        {
+          "orth": "ひ",
+          "pron": "ヒ",
+          "pron_base": "ヒ",
+          "recorded_aType": 2,
+          "tokenizer_moras_pron": 1,
+          "tokenizer_moras_lemma": 1,
+          "pos": "名詞-数詞"
+        },
+        {
+          "orth": "ふ",
+          "pron": "フ",
+          "pron_base": "フ",
+          "recorded_aType": 2,
+          "tokenizer_moras_pron": 1,
+          "tokenizer_moras_lemma": 1,
+          "pos": "名詞-数詞"
+        },
+        {
+          "orth": "ぽおん",
+          "pron": "ポーン",
+          "pron_base": "ポーン",
+          "recorded_aType": 4,
+          "tokenizer_moras_pron": 3,
+          "tokenizer_moras_lemma": 3,
+          "pos": "副詞-*"
+        },
+        {
+          "orth": "ぽーん",
+          "pron": "ポーン",
+          "pron_base": "ポーン",
+          "recorded_aType": 4,
+          "tokenizer_moras_pron": 3,
+          "tokenizer_moras_lemma": 3,
+          "pos": "副詞-*"
+        },
+        {
+          "orth": "アテュス",
+          "pron": "アチュス",
+          "pron_base": "アチュス",
+          "recorded_aType": 10,
+          "tokenizer_moras_pron": 3,
+          "tokenizer_moras_lemma": 3,
+          "pos": "名詞-固有名詞"
+        },
+        {
+          "orth": "ウヂノワキ",
+          "pron": "ウジノワキ",
+          "pron_base": "ウジノワキ",
+          "recorded_aType": 7,
+          "tokenizer_moras_pron": 5,
+          "tokenizer_moras_lemma": 5,
+          "pos": "名詞-固有名詞"
+        },
+        {
+          "orth": "エレクトロ",
+          "pron": "エレクトロ",
+          "pron_base": "エレクトロ",
+          "recorded_aType": 8,
+          "tokenizer_moras_pron": 5,
+          "tokenizer_moras_lemma": 5,
+          "pos": "名詞-普通名詞"
+        },
+        {
+          "orth": "サラン",
+          "pron": "サラン",
+          "pron_base": "サラン",
+          "recorded_aType": 4,
+          "tokenizer_moras_pron": 3,
+          "tokenizer_moras_lemma": 3,
+          "pos": "名詞-固有名詞"
+        },
+        {
+          "orth": "スマ",
+          "pron": "スマ",
+          "pron_base": "スマ",
+          "recorded_aType": 4,
+          "tokenizer_moras_pron": 2,
+          "tokenizer_moras_lemma": 2,
+          "pos": "名詞-普通名詞"
+        },
+        {
+          "orth": "スルホ",
+          "pron": "スルホ",
+          "pron_base": "スルホ",
+          "recorded_aType": 4,
+          "tokenizer_moras_pron": 3,
+          "tokenizer_moras_lemma": 3,
+          "pos": "名詞-普通名詞"
+        },
+        {
+          "orth": "トム",
+          "pron": "トム",
+          "pron_base": "トム",
+          "recorded_aType": 6,
+          "tokenizer_moras_pron": 2,
+          "tokenizer_moras_lemma": 2,
+          "pos": "名詞-普通名詞"
+        },
+        {
+          "orth": "ドゥバイ",
+          "pron": "ドゥバイ",
+          "pron_base": "ドゥバイ",
+          "recorded_aType": 10,
+          "tokenizer_moras_pron": 3,
+          "tokenizer_moras_lemma": 3,
+          "pos": "名詞-固有名詞"
+        },
+        {
+          "orth": "バンデグラフ",
+          "pron": "バンデグラフ",
+          "pron_base": "バンデグラフ",
+          "recorded_aType": 7,
+          "tokenizer_moras_pron": 6,
+          "tokenizer_moras_lemma": 6,
+          "pos": "名詞-普通名詞"
+        },
+        {
+          "orth": "フェデラル",
+          "pron": "フェデラル",
+          "pron_base": "フェデラル",
+          "recorded_aType": 5,
+          "tokenizer_moras_pron": 4,
+          "tokenizer_moras_lemma": 4,
+          "pos": "名詞-普通名詞"
+        },
+        {
+          "orth": "プレスト",
+          "pron": "プレスト",
+          "pron_base": "プレスト",
+          "recorded_aType": 5,
+          "tokenizer_moras_pron": 4,
+          "tokenizer_moras_lemma": 4,
+          "pos": "名詞-普通名詞"
+        },
+        {
+          "orth": "ペブル",
+          "pron": "ペブル",
+          "pron_base": "ペブル",
+          "recorded_aType": 4,
+          "tokenizer_moras_pron": 3,
+          "tokenizer_moras_lemma": 3,
+          "pos": "名詞-固有名詞"
+        },
+        {
+          "orth": "ペント",
+          "pron": "ペント",
+          "pron_base": "ペント",
+          "recorded_aType": 4,
+          "tokenizer_moras_pron": 3,
+          "tokenizer_moras_lemma": 3,
+          "pos": "名詞-普通名詞"
+        },
+        {
+          "orth": "ポーン",
+          "pron": "ポーン",
+          "pron_base": "ポーン",
+          "recorded_aType": 4,
+          "tokenizer_moras_pron": 3,
+          "tokenizer_moras_lemma": 3,
+          "pos": "副詞-*"
+        },
+        {
+          "orth": "マーシュ",
+          "pron": "マーシュ",
+          "pron_base": "マーシュ",
+          "recorded_aType": 11,
+          "tokenizer_moras_pron": 3,
+          "tokenizer_moras_lemma": 3,
+          "pos": "名詞-固有名詞"
+        },
+        {
+          "orth": "ミクー",
+          "pron": "ミクー",
+          "pron_base": "ミクー",
+          "recorded_aType": 10,
+          "tokenizer_moras_pron": 3,
+          "tokenizer_moras_lemma": 3,
+          "pos": "名詞-固有名詞"
+        },
+        {
+          "orth": "ラオアグ",
+          "pron": "ラオアグ",
+          "pron_base": "ラオアグ",
+          "recorded_aType": 7,
+          "tokenizer_moras_pron": 4,
+          "tokenizer_moras_lemma": 4,
+          "pos": "名詞-固有名詞"
+        },
+        {
+          "orth": "リンチ",
+          "pron": "リンチ",
+          "pron_base": "リンチ",
+          "recorded_aType": 4,
+          "tokenizer_moras_pron": 3,
+          "tokenizer_moras_lemma": 3,
+          "pos": "名詞-固有名詞"
+        },
+        {
+          "orth": "レッセ",
+          "pron": "レッセ",
+          "pron_base": "レッセ",
+          "recorded_aType": 4,
+          "tokenizer_moras_pron": 3,
+          "tokenizer_moras_lemma": 3,
+          "pos": "名詞-普通名詞"
+        },
+        {
+          "orth": "ワクスマン",
+          "pron": "ワクスマン",
+          "pron_base": "ワクスマン",
+          "recorded_aType": 11,
+          "tokenizer_moras_pron": 5,
+          "tokenizer_moras_lemma": 5,
+          "pos": "名詞-固有名詞"
+        }
+      ]
+    }
+  }
+}

stats.json

@@ -0,0 +1,60 @@
+{
+  "built_at_seconds": 24.3,
+  "n_total": 89539,
+  "by_split": {
+    "train": 71544,
+    "test": 9036,
+    "dev": 8959
+  },
+  "n_hard_in_test": 818,
+  "n_hard_all_in_test": 105,
+  "hard_fraction_of_test": 0.0905,
+  "n_with_accent_variation": 1030,
+  "by_script": {
+    "mixed": 27290,
+    "katakana": 7012,
+    "kanji": 55237
+  },
+  "by_goshu": {
+    "外": 4443,
+    "漢": 23313,
+    "和": 34760,
+    "固": 25085,
+    "None": 764,
+    "混": 604,
+    "記号": 570
+  },
+  "by_n_components": {
+    "2": 70419,
+    "3": 16286,
+    "4": 2539,
+    "5": 275,
+    "6": 19,
+    "7": 1
+  },
+  "by_n_moras": {
+    "6": 18884,
+    "4": 21228,
+    "5": 19522,
+    "7": 11112,
+    "11": 2003,
+    "9": 3836,
+    "8": 7419,
+    "10": 2958,
+    "12": 1298,
+    "13": 593,
+    "14": 390,
+    "15": 198,
+    "16": 90,
+    "17": 7,
+    "18": 1
+  },
+  "n_proper": 59430,
+  "n_common": 30109,
+  "n_all_components_in_lexicon": 89539,
+  "sources": {
+    "naist-jdic": "BSD 3-clause, Nara Institute of Science and Technology, via pyopenjtalk",
+    "unidic-2.1.2": "GPL/LGPL/BSD-3 tri-licence (BSD option used), The UniDic Consortium (NINJAL)"
+  },
+  "note": "gold_accents is a SET: >1 entry means the dictionaries attest more than one accent. No human annotation is present in this release."
+}

train.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff2791d0940864bca0d07b7682e767297453a376b418e45300e56601fe3b640d
+size 52538706