Update fixes, candidates and predictions by segment

README.md

@@ -0,0 +1,105 @@
+---
+license: cc-by-4.0
+language:
+- en
+- he
+pretty_name: ASR Error Prediction (Hebrish)
+tags:
+- asr
+- speech-recognition
+- error-prediction
+- code-switching
+- fine-tuning
+- hebrish
+task_categories:
+- text-classification
+size_categories:
+- 1K<n<10K
+configs:
+- config_name: fixes
+  data_files:
+  - split: codeswitching
+    path: fixes/codeswitching.jsonl
+  - split: tech
+    path: fixes/tech.jsonl
+  - split: names
+    path: fixes/names.jsonl
+  - split: general_english
+    path: fixes/general_english.jsonl
+- config_name: candidates
+  data_files:
+  - split: codeswitching
+    path: candidates/codeswitching.jsonl
+  - split: tech
+    path: candidates/tech.jsonl
+- config_name: predictions-haiku-r1
+  data_files:
+  - split: codeswitching
+    path: predictions/haiku-r1/codeswitching.jsonl
+  - split: tech
+    path: predictions/haiku-r1/tech.jsonl
+  - split: general_english
+    path: predictions/haiku-r1/general_english.jsonl
+---
+
+# ASR Error Prediction
+
+Training data for a model that predicts **which words a speech recognition model will get wrong for
+one speaker**, by extrapolating from the words it has already been confirmed to get wrong. Built for
+the [Hebrish](https://hebrish.app) fine-tunes of `nvidia/parakeet-tdt-0.6b-v2` (English with
+romanized Hebrew code-switching). Plan, scripts and method:
+[danielrosehill/ASR-Mistranscription-Predictor](https://github.com/danielrosehill/ASR-Mistranscription-Predictor).
+
+## Segments
+
+Every config is split by **segment**, so predictions can be made and used per training dataset:
+
+| Segment | What it covers | Source of candidates |
+|---|---|---|
+| `codeswitching` | romanized Hebrew and Yiddish terms | the Hebrish lexicon (places, food, bureaucracy, religious life, ...) |
+| `tech` | tech, AI, Linux and smart-home terms and acronyms | the project's tech vocabulary packs, plus LLM proposals |
+| `names` | people, companies, projects | none yet (fixes only) |
+| `general_english` | ordinary English words | LLM proposals |
+
+| Config | codeswitching | tech | names | general_english |
+|---|---|---|---|---|
+| `fixes` | 7 | 59 | 9 | 12 |
+| `candidates` | 3634 | 1112 | 0 | 0 |
+| `predictions-haiku-r1` | 3634 | 1163 | 0 | 55 |
+
+## Configs
+
+- **`fixes`**: the positives. Words confirmed wrong in the speaker's real dictation, from
+  [`Hebrish/ASR-Error-List`](https://huggingface.co/datasets/Hebrish/ASR-Error-List) (pinned in
+  `error_list_rev`): `word`, `heard_as` (the wrong forms the ASR wrote), `kinds`, `occurrences`,
+  `models`, dates, and `label` (`acoustic`, or `orthographic` when the only error is casing or a
+  hyphen).
+- **`candidates`**: words the speaker is likely to say that are not on the fix list: `term`,
+  `domain`, `meaning`, `takes` (recorded training takes so far; null = unknown) and `pool` (source
+  file and commit).
+- **`predictions-<run>`**: one model's scores for the candidates, plus terms it proposed outside the
+  pools (`origin`: `pool` or `proposed`): `p_fail` (0-100), `pattern` (the confusion class it
+  applied), `likely_heard_as`, `reason`, `rank` within the segment.
+
+**Predictions are not labels.** They are rankings to be checked against the real model (a synthetic
+probe, then a few real recordings) before any word gets training sentences. Confirmed ones move to
+the error list with `source: predicted`.
+
+## Runs
+
+| Run | Date | Model | Notes |
+|---|---|---|---|
+| `haiku-r1` | 2026-10-25 | `claude-haiku-5-5` (Anthropic Batch API, low effort; batch `msgbatch_01SFKjYbEtvGqChr8eo3JXPk`, US$0.13) | first pass; fix list at 87 words; scores every candidate in chunks of 60 with the whole fix list as evidence, plus 52 proposed tech terms and 55 proposed English words; 1 candidate missing from the replies |
+
+**`haiku-r1` caveats.** Scores are uncalibrated and not comparable across segments. Code-switching
+scores run high (mean 58 for terms with no recorded takes, 37% of candidates at 60 or above) because
+the fix list has only 7 Hebrew words to learn from, so Haiku leans on "Hebrew = hard". Scores do
+fall with recorded takes (mean 58 at 0 takes, 49 at 1 to 4, 31 at 5 to 19, 22 at 20 or more), which
+is the expected direction. Use ranks within a segment, not thresholds. Wrong forms that equalled
+the term itself (1,175 of them) were removed after the run.
+
+## Not included
+
+No audio, and nothing from the speaker's dictations beyond single confirmed words and the ASR's
+wrong forms. Negatives (rare words the model transcribed correctly) are planned but not published
+yet.

fixes/codeswitching.jsonl

@@ -0,0 +1,7 @@
+{"word": "Netanyahu", "segment": "codeswitching", "heard_as": ["Etanyahu", "Etnia"], "kinds": ["name", "hebrew"], "meaning": "Israeli prime minister", "occurrences": 2, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-22", "last_seen": "2026-10-22", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Bezeq", "segment": "codeswitching", "heard_as": ["Bezek"], "kinds": ["hebrew", "name"], "meaning": "Israeli telecom / internet provider", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-25", "last_seen": "2026-10-25", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Bituach Leumi", "segment": "codeswitching", "heard_as": ["betuachlomi"], "kinds": ["hebrew"], "meaning": "National Insurance Institute", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-23", "last_seen": "2026-10-23", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Chevrat Hashmal", "segment": "codeswitching", "heard_as": ["Chashmal Yisrael"], "kinds": ["hebrew", "name"], "meaning": "Israel Electric Corporation", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-25", "last_seen": "2026-10-25", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Hagihon", "segment": "codeswitching", "heard_as": ["Gichon"], "kinds": ["hebrew", "name"], "meaning": "Jerusalem water corporation", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-25", "last_seen": "2026-10-25", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Meorav", "segment": "codeswitching", "heard_as": ["Maorab"], "kinds": ["personal", "hebrew"], "meaning": "Daniel's TTS voice-clone model; Hebrew for 'mixed'", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-22", "last_seen": "2026-10-22", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Sabich", "segment": "codeswitching", "heard_as": ["Tzabich"], "kinds": ["personal", "hebrew"], "meaning": "Daniel's main ASR fine-tune, named after the Israeli sandwich", "occurrences": 1, "models": [], "first_seen": "2026-10-21", "last_seen": "2026-10-21", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}

fixes/general_english.jsonl

@@ -0,0 +1,12 @@
+{"word": "ab initio", "segment": "general_english", "heard_as": ["a benissio", "abenisio"], "kinds": ["other"], "meaning": "Latin: from the beginning", "occurrences": 2, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "series", "segment": "general_english", "heard_as": ["Ceres", "Siries"], "kinds": ["other"], "meaning": null, "occurrences": 2, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "do you want me", "segment": "general_english", "heard_as": ["Dramat me"], "kinds": ["other"], "meaning": "Plain English phrase", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-25", "last_seen": "2026-10-25", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "inbound", "segment": "general_english", "heard_as": ["for bound"], "kinds": ["other"], "meaning": "Incoming (email)", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-25", "last_seen": "2026-10-25", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Kaizen", "segment": "general_english", "heard_as": ["Kaiser"], "kinds": ["other"], "meaning": "continuous small-step improvement", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "lav mic", "segment": "general_english", "heard_as": ["Lav Mike"], "kinds": ["other"], "meaning": "Lavalier microphone", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-25", "last_seen": "2026-10-25", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "marking", "segment": "general_english", "heard_as": ["marketing"], "kinds": ["other"], "meaning": "As in marking complete", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "populated", "segment": "general_english", "heard_as": ["opulated"], "kinds": ["other"], "meaning": "Filled in", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-25", "last_seen": "2026-10-25", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "prosody", "segment": "general_english", "heard_as": ["porosity"], "kinds": ["other"], "meaning": "the rhythm, stress and intonation of speech, as in TTS", "occurrences": 1, "models": ["shakshuka Shakshuka v2.0"], "first_seen": "2026-10-22", "last_seen": "2026-10-22", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "subscription", "segment": "general_english", "heard_as": ["prescription"], "kinds": ["other"], "meaning": "Subscription (Claude Max)", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-25", "last_seen": "2026-10-25", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "tell her", "segment": "general_english", "heard_as": ["Teller"], "kinds": ["other"], "meaning": "Plain English phrase", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-25", "last_seen": "2026-10-25", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "vocab", "segment": "general_english", "heard_as": ["Rocab"], "kinds": ["other"], "meaning": "Vocabulary", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}

fixes/names.jsonl

@@ -0,0 +1,9 @@
+{"word": "Corn", "segment": "names", "heard_as": ["Coren", "Cornak's", "Eminent Corn", "Corner", "Korn", "Corin"], "kinds": ["personal", "name"], "meaning": "co-host on Daniel's podcast My Weird Prompts", "occurrences": 6, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-23", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "My Weird Prompts", "segment": "names", "heard_as": ["MyWord Prompts", "My WordPrompts", "My Word Promise", "mywordprompt"], "kinds": ["personal"], "meaning": "Daniel's AI podcast", "occurrences": 6, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-22", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Boxty", "segment": "names", "heard_as": ["E box d"], "kinds": ["personal"], "meaning": null, "occurrences": 1, "models": [], "first_seen": "2026-10-21", "last_seen": "2026-10-21", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Carrot Cake AI", "segment": "names", "heard_as": ["Karrot Cake AI"], "kinds": ["personal"], "meaning": "a company/brand name Daniel uses, linked from his profiles", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-22", "last_seen": "2026-10-22", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "DSR Holdings", "segment": "names", "heard_as": ["Vidyester Holdings"], "kinds": ["personal", "name"], "meaning": "Daniel's company", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-23", "last_seen": "2026-10-23", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Ezra", "segment": "names", "heard_as": ["Esther"], "kinds": ["personal", "name"], "meaning": "the speaker's baby son, a toddler born in 2025; sentences about family life, never work", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Hegseth", "segment": "names", "heard_as": ["Hagsets"], "kinds": ["name"], "meaning": "US Secretary of Defense", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-23", "last_seen": "2026-10-23", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Janglo", "segment": "names", "heard_as": ["Danglo"], "kinds": ["name"], "meaning": "English-speakers' community site in Israel", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-23", "last_seen": "2026-10-23", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "pseudospam", "segment": "names", "heard_as": ["Surasfam"], "kinds": ["personal"], "meaning": "Daniel's term for generic non-spam bulk email", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-25", "last_seen": "2026-10-25", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}

fixes/tech.jsonl

@@ -0,0 +1,59 @@
+{"word": "ROCm", "segment": "tech", "heard_as": ["ROCM", "Rockham", "Rockin", "RCM", "Rock'em"], "kinds": ["tech", "linux"], "meaning": "AMD's open-source GPU compute stack (like CUDA)", "occurrences": 15, "models": ["sabich Sabich v2.0", "shakshuka Shakshuka v2.0"], "first_seen": "2026-10-21", "last_seen": "2026-10-23", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "LoRA", "segment": "tech", "heard_as": ["Lores", "Loas", "Laura", "lores", "Loa", "LoRa", "lore", "lures"], "kinds": ["tech", "ai_interactions"], "meaning": "low-rank adapter fine-tuning", "occurrences": 14, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-23", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Claude Code", "segment": "tech", "heard_as": ["clawed code", "clawed", "claud", "clawed coat", "claw code", "Core Code", "claudcode", "claud Code", "the claud", "clause claud code"], "kinds": ["tech", "ai_interactions"], "meaning": "Anthropic's coding agent CLI", "occurrences": 12, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-22", "last_seen": "2026-10-25", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Hugging Face", "segment": "tech", "heard_as": ["Hooking Face", "hugging face", "hooking face", "Hogging Face", "hogging face"], "kinds": ["tech", "ai_interactions"], "meaning": "the AI model and dataset hub", "occurrences": 6, "models": ["sabich Sabich v2.0", "shakshuka Shakshuka v2.0"], "first_seen": "2026-10-21", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Hebrish", "segment": "tech", "heard_as": ["HeBrish", "Hebris", "keybrush", "Hebrew Shop", "Hebrew shtal"], "kinds": ["personal", "tech"], "meaning": "Daniel's Hebrew-English code-switching ASR project, app and Hugging Face org", "occurrences": 5, "models": ["sabich Sabich v2.0", "shakshuka Shakshuka v2.0"], "first_seen": "2026-10-21", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Typst", "segment": "tech", "heard_as": ["TyPst", "TIPS-type", "tips toypt", "TyPstip", "tipst"], "kinds": ["tech"], "meaning": "Markup-based typesetting system", "occurrences": 5, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-25", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "NocoDB", "segment": "tech", "heard_as": ["No CodeDB", "Nocode", "NokoDB", "Noko DB"], "kinds": ["tech"], "meaning": "open-source no-code database", "occurrences": 4, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-23", "last_seen": "2026-10-23", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "CLAUDE.md", "segment": "tech", "heard_as": ["userlevelclaw.md", "4 MD", "Claude NMD"], "kinds": ["ai_interactions", "tech"], "meaning": "Claude Code instructions file", "occurrences": 3, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-25", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Cloudflare", "segment": "tech", "heard_as": ["cloud fair", "Cloud Claire", "Cloud Clare"], "kinds": ["tech"], "meaning": "CDN and network provider", "occurrences": 3, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-23", "last_seen": "2026-10-23", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Colab", "segment": "tech", "heard_as": ["Collab", "collab"], "kinds": ["tech"], "meaning": "Google's hosted notebook service", "occurrences": 3, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-23", "last_seen": "2026-10-23", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "F-Droid", "segment": "tech", "heard_as": ["FDroid", "Fdroid"], "kinds": ["tech"], "meaning": "the open-source Android app store", "occurrences": 3, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-21", "last_seen": "2026-10-23", "label": "orthographic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Nemotron", "segment": "tech", "heard_as": ["Nematron", "Nementron"], "kinds": ["tech", "ai_interactions"], "meaning": "NVIDIA's Nemotron model family (a streaming ASR model here)", "occurrences": 3, "models": ["sabich Sabich v2.0", "shakshuka Shakshuka v2.0"], "first_seen": "2026-10-22", "last_seen": "2026-10-22", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "skill", "segment": "tech", "heard_as": ["scale"], "kinds": ["ai_interactions"], "meaning": "Agent skill (Claude Code)", "occurrences": 3, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-25", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Unsloth", "segment": "tech", "heard_as": ["Unsoft", "On Sloth", "Onsloth"], "kinds": ["tech", "ai_interactions"], "meaning": "fine-tuning library and platform", "occurrences": 3, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-23", "last_seen": "2026-10-23", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "DeepSeek", "segment": "tech", "heard_as": ["DeepSee", "DeepC"], "kinds": ["ai_interactions", "tech"], "meaning": "Chinese LLM provider", "occurrences": 2, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-25", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "disfluency", "segment": "tech", "heard_as": ["disluency", "Dispuency"], "kinds": ["tech"], "meaning": "fillers and false starts in speech", "occurrences": 2, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "DMARC", "segment": "tech", "heard_as": ["GMARC", "Dimark"], "kinds": ["tech"], "meaning": "Email authentication policy", "occurrences": 2, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-23", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "evals", "segment": "tech", "heard_as": ["evalves", "valves"], "kinds": ["tech", "ai_interactions"], "meaning": "AI model evaluations", "occurrences": 2, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-23", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Infisical", "segment": "tech", "heard_as": ["Infiniscal", "Infinisco"], "kinds": ["tech"], "meaning": "secrets manager", "occurrences": 2, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-23", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Jupyter", "segment": "tech", "heard_as": ["Jupiter"], "kinds": ["tech"], "meaning": "Open-source project behind Jupyter Notebook and JupyterLab; pronounced like the planet Jupiter", "occurrences": 2, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-22", "last_seen": "2026-10-23", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Karpathy", "segment": "tech", "heard_as": ["Karpathi", "Karpathi's"], "kinds": ["name", "tech"], "meaning": "Andrej Karpathy, AI researcher and educator (ex-OpenAI, ex-Tesla AI)", "occurrences": 2, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-22", "last_seen": "2026-10-22", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "ONNX", "segment": "tech", "heard_as": ["ONIX", "onyx"], "kinds": ["tech", "ai_interactions"], "meaning": "the Open Neural Network Exchange model format", "occurrences": 2, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-21", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Qwen", "segment": "tech", "heard_as": ["Qwen1", "quen"], "kinds": ["tech", "ai_interactions"], "meaning": "Alibaba's Qwen model family", "occurrences": 2, "models": ["sabich Sabich v2.0", "shakshuka Shakshuka v2.0"], "first_seen": "2026-10-22", "last_seen": "2026-10-23", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Z alias", "segment": "tech", "heard_as": ["ZALAS", "Zed Elias"], "kinds": ["linux"], "meaning": "zoxide shortcut to a repo", "occurrences": 2, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "abliteration", "segment": "tech", "heard_as": ["Obliteration"], "kinds": ["tech", "ai_interactions"], "meaning": "removing refusal behaviour from model weights", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-23", "last_seen": "2026-10-23", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "agentic", "segment": "tech", "heard_as": ["agenda"], "kinds": ["ai_interactions"], "meaning": null, "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "AMD", "segment": "tech", "heard_as": ["MD"], "kinds": ["tech"], "meaning": "GPU and CPU maker", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-23", "last_seen": "2026-10-23", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "APK", "segment": "tech", "heard_as": ["ABK"], "kinds": ["tech"], "meaning": "Android app package", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-23", "last_seen": "2026-10-23", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "ASR", "segment": "tech", "heard_as": ["ACR"], "kinds": ["tech"], "meaning": "automatic speech recognition", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Chatterbox", "segment": "tech", "heard_as": ["Shatterbox"], "kinds": ["tech", "ai_interactions"], "meaning": "Resemble AI's open-source TTS model", "occurrences": 1, "models": ["shakshuka Shakshuka v2.0"], "first_seen": "2026-10-22", "last_seen": "2026-10-22", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Claude in Chrome", "segment": "tech", "heard_as": ["Claude Grouping Chrome"], "kinds": ["ai_interactions"], "meaning": "Claude browser extension", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-25", "last_seen": "2026-10-25", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "co.il", "segment": "tech", "heard_as": ["co.al"], "kinds": ["tech"], "meaning": "Israeli domain suffix", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-25", "last_seen": "2026-10-25", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "eSpeak", "segment": "tech", "heard_as": ["Espeak"], "kinds": ["linux"], "meaning": "the open-source speech synthesizer", "occurrences": 1, "models": ["shakshuka Shakshuka v2.0"], "first_seen": "2026-10-22", "last_seen": "2026-10-22", "label": "orthographic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "EXE", "segment": "tech", "heard_as": ["AXE"], "kinds": ["tech"], "meaning": "the Windows executable file format, said as letters", "occurrences": 1, "models": [], "first_seen": "2026-10-21", "last_seen": "2026-10-21", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "fal", "segment": "tech", "heard_as": ["Fel"], "kinds": ["tech", "ai_interactions"], "meaning": "generative media API platform (fal.ai)", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "few-shot", "segment": "tech", "heard_as": ["fusil"], "kinds": ["ai_interactions"], "meaning": "prompting with a few examples", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-23", "last_seen": "2026-10-23", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Jupyter Notebook", "segment": "tech", "heard_as": ["Jupiter notebooks"], "kinds": ["tech"], "meaning": "Interactive Python notebook environment (.ipynb files) for data science and ML", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-22", "last_seen": "2026-10-22", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Kdenlive", "segment": "tech", "heard_as": ["Caden Live"], "kinds": ["linux"], "meaning": "the KDE video editor", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-22", "last_seen": "2026-10-22", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "LLM", "segment": "tech", "heard_as": ["nelam"], "kinds": ["ai_interactions"], "meaning": "large language model", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Muse Spark", "segment": "tech", "heard_as": ["News Spark"], "kinds": ["ai_interactions"], "meaning": "LLM", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Nvidia", "segment": "tech", "heard_as": ["NVideo"], "kinds": ["tech"], "meaning": "GPU maker", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-22", "last_seen": "2026-10-22", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Ollama", "segment": "tech", "heard_as": ["ulama"], "kinds": ["tech", "ai_interactions"], "meaning": "local LLM runner", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "OPNsense", "segment": "tech", "heard_as": ["OpenSense"], "kinds": ["tech", "linux"], "meaning": "Open-source firewall OS", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Opus", "segment": "tech", "heard_as": ["Focus"], "kinds": ["ai_interactions"], "meaning": "Claude Opus model, as in Opus 5.5", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-25", "last_seen": "2026-10-25", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "post-processing", "segment": "tech", "heard_as": ["PostgreSQL processing"], "kinds": ["tech"], "meaning": "cleanup applied to ASR output", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Qdrant", "segment": "tech", "heard_as": ["Qdrint"], "kinds": ["tech"], "meaning": "open-source vector database", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "rclone", "segment": "tech", "heard_as": ["Arclone"], "kinds": ["tech", "linux"], "meaning": "Cloud storage sync CLI", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-25", "last_seen": "2026-10-25", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "regex", "segment": "tech", "heard_as": ["RedJack"], "kinds": ["tech"], "meaning": "regular expression", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-23", "last_seen": "2026-10-23", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "remote", "segment": "tech", "heard_as": ["ote"], "kinds": ["tech"], "meaning": "As in remote agent", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-25", "last_seen": "2026-10-25", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "repo", "segment": "tech", "heard_as": ["Refo"], "kinds": ["tech"], "meaning": "Repository", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "SaaS", "segment": "tech", "heard_as": ["SAS"], "kinds": ["tech"], "meaning": "software as a service", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-23", "last_seen": "2026-10-23", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "SBC", "segment": "tech", "heard_as": ["SPCAdmin"], "kinds": ["tech", "linux"], "meaning": "Single-board computer, e.g. Raspberry Pi", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Sonnet", "segment": "tech", "heard_as": ["Sommet"], "kinds": ["tech", "ai_interactions"], "meaning": "Anthropic's mid-tier Claude model", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-23", "last_seen": "2026-10-23", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "SSE", "segment": "tech", "heard_as": ["SSC"], "kinds": ["tech"], "meaning": "server-sent events (MCP transport)", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Superwhisper", "segment": "tech", "heard_as": ["Superwisper"], "kinds": ["tech"], "meaning": "macOS dictation app", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-24", "last_seen": "2026-10-24", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "system tray", "segment": "tech", "heard_as": ["system try"], "kinds": ["linux"], "meaning": "Desktop notification area", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-25", "last_seen": "2026-10-25", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "tab", "segment": "tech", "heard_as": ["top"], "kinds": ["tech"], "meaning": "Browser tab", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-25", "last_seen": "2026-10-25", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "TTS", "segment": "tech", "heard_as": ["CTS"], "kinds": ["tech", "ai_interactions"], "meaning": "text-to-speech", "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-23", "last_seen": "2026-10-23", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}
+{"word": "Whisperian", "segment": "tech", "heard_as": ["Wisperion"], "kinds": ["tech"], "meaning": null, "occurrences": 1, "models": ["sabich Sabich v2.0"], "first_seen": "2026-10-22", "last_seen": "2026-10-22", "label": "acoustic", "source": "dictation", "error_list_rev": "e325fd8180b3"}

predictions/haiku-r1/general_english.jsonl

@@ -0,0 +1,55 @@
+{"term": "ephemeral", "segment": "general_english", "domain": null, "takes": null, "p_fail": 45, "pattern": "snap_to_neighbour", "likely_heard_as": ["ephemeral's", "ethereal", "femoral"], "reason": "Rare word with near-neighbour English words sharing vowels and endings.", "origin": "proposed", "rank": 1, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "hors d'oeuvre", "segment": "general_english", "domain": null, "takes": null, "p_fail": 45, "pattern": "word_boundary", "likely_heard_as": ["or dover", "or doeuvre"], "reason": "Foreign phrase split and spelled variously.", "origin": "proposed", "rank": 2, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "epitome", "segment": "general_english", "domain": null, "takes": null, "p_fail": 40, "pattern": "snap_to_neighbour", "likely_heard_as": ["epitomy", "anatomy"], "reason": "Uncommon word resembling more frequent words; similar to porosity-for-prosody.", "origin": "proposed", "rank": 3, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "prerequisite", "segment": "general_english", "domain": null, "takes": null, "p_fail": 40, "pattern": "word_boundary", "likely_heard_as": ["pre requisite", "pre-requisite"], "reason": "Compound-like word often split or hyphenated wrongly.", "origin": "proposed", "rank": 4, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "cacophony", "segment": "general_english", "domain": null, "takes": null, "p_fail": 40, "pattern": "snap_to_neighbour", "likely_heard_as": ["cacophonic", "catastrophe"], "reason": "Rare word with near-neighbour forms.", "origin": "proposed", "rank": 5, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "onerous", "segment": "general_english", "domain": null, "takes": null, "p_fail": 40, "pattern": "snap_to_neighbour", "likely_heard_as": ["onerus", "honorous"], "reason": "Rare word close to common sounds.", "origin": "proposed", "rank": 6, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "reticent", "segment": "general_english", "domain": null, "takes": null, "p_fail": 40, "pattern": "snap_to_neighbour", "likely_heard_as": ["reticence", "recent"], "reason": "Rare word near frequent recent.", "origin": "proposed", "rank": 7, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "ubiquitous", "segment": "general_english", "domain": null, "takes": null, "p_fail": 40, "pattern": "snap_to_neighbour", "likely_heard_as": ["ubiquitus", "ubiquity"], "reason": "Rare word with stress and vowel variation.", "origin": "proposed", "rank": 8, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "ptarmigan", "segment": "general_english", "domain": null, "takes": null, "p_fail": 40, "pattern": "initial_drop", "likely_heard_as": ["tarmigan", "tar migan"], "reason": "Silent initial p dropped and rare word.", "origin": "proposed", "rank": 9, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "hierarchy", "segment": "general_english", "domain": null, "takes": null, "p_fail": 35, "pattern": "snap_to_neighbour", "likely_heard_as": ["hierarchy's", "heretic"], "reason": "Multisyllabic word with stress shift; snaps to similar-sounding frequent forms.", "origin": "proposed", "rank": 10, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "sceptical", "segment": "general_english", "domain": null, "takes": null, "p_fail": 35, "pattern": "orthographic", "likely_heard_as": ["skeptical", "sceptic al"], "reason": "Spelling variant; model may choose the other orthography.", "origin": "proposed", "rank": 11, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "inadvertently", "segment": "general_english", "domain": null, "takes": null, "p_fail": 35, "pattern": "snap_to_neighbour", "likely_heard_as": ["inadvertent", "in advertently"], "reason": "Long adverb that may be split or truncated.", "origin": "proposed", "rank": 12, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "anecdote", "segment": "general_english", "domain": null, "takes": null, "p_fail": 35, "pattern": "snap_to_neighbour", "likely_heard_as": ["antidote", "anecdotes"], "reason": "Near-homophone neighbour antidote shares most sounds.", "origin": "proposed", "rank": 13, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "meticulous", "segment": "general_english", "domain": null, "takes": null, "p_fail": 35, "pattern": "snap_to_neighbour", "likely_heard_as": ["meticulously", "metaculous"], "reason": "Rare word with stress and vowel variation.", "origin": "proposed", "rank": 14, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "hypothesis", "segment": "general_english", "domain": null, "takes": null, "p_fail": 30, "pattern": "snap_to_neighbour", "likely_heard_as": ["hypothesize", "hypothetical"], "reason": "Common confusion with morphologically related forms.", "origin": "proposed", "rank": 15, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "nevertheless", "segment": "general_english", "domain": null, "takes": null, "p_fail": 30, "pattern": "word_boundary", "likely_heard_as": ["never the less", "nevertheless,"], "reason": "Long function word often split into two words.", "origin": "proposed", "rank": 16, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "cappuccino", "segment": "general_english", "domain": null, "takes": null, "p_fail": 30, "pattern": "consonant_shift", "likely_heard_as": ["capuccino", "cappucino"], "reason": "Double consonant spelling variants.", "origin": "proposed", "rank": 17, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "vicinity", "segment": "general_english", "domain": null, "takes": null, "p_fail": 30, "pattern": "snap_to_neighbour", "likely_heard_as": ["vicinities", "vicinity's"], "reason": "Uncommon noun with plural drift.", "origin": "proposed", "rank": 18, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "autonomy", "segment": "general_english", "domain": null, "takes": null, "p_fail": 30, "pattern": "snap_to_neighbour", "likely_heard_as": ["autonomous", "autonomy's"], "reason": "Similar derivational forms.", "origin": "proposed", "rank": 19, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "pedantic", "segment": "general_english", "domain": null, "takes": null, "p_fail": 30, "pattern": "snap_to_neighbour", "likely_heard_as": ["pedantics", "pedantically"], "reason": "Unusual word with suffix variation.", "origin": "proposed", "rank": 20, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "pneumonia", "segment": "general_english", "domain": null, "takes": null, "p_fail": 30, "pattern": "initial_drop", "likely_heard_as": ["numonia", "new monia"], "reason": "Silent initial p dropped.", "origin": "proposed", "rank": 21, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "licorice", "segment": "general_english", "domain": null, "takes": null, "p_fail": 30, "pattern": "orthographic", "likely_heard_as": ["liquorice", "licorise"], "reason": "Spelling variant.", "origin": "proposed", "rank": 22, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "dilemma", "segment": "general_english", "domain": null, "takes": null, "p_fail": 25, "pattern": "snap_to_neighbour", "likely_heard_as": ["dilemmas", "delmar"], "reason": "Unusual stress pattern with possible plural or name-like output.", "origin": "proposed", "rank": 23, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "consensus", "segment": "general_english", "domain": null, "takes": null, "p_fail": 25, "pattern": "snap_to_neighbour", "likely_heard_as": ["consensus's", "consent sus"], "reason": "Word boundary and final cluster may be misheard.", "origin": "proposed", "rank": 24, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "quantum", "segment": "general_english", "domain": null, "takes": null, "p_fail": 25, "pattern": "snap_to_neighbour", "likely_heard_as": ["quantom", "quanta"], "reason": "Vowel and consonant variation in spelling.", "origin": "proposed", "rank": 25, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "lasagna", "segment": "general_english", "domain": null, "takes": null, "p_fail": 25, "pattern": "orthographic", "likely_heard_as": ["lasagne", "lasania"], "reason": "Spelling variants of foreign loanword.", "origin": "proposed", "rank": 26, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "mischievous", "segment": "general_english", "domain": null, "takes": null, "p_fail": 25, "pattern": "snap_to_neighbour", "likely_heard_as": ["mischievious", "mischief us"], "reason": "Spelling and syllable reduction issues.", "origin": "proposed", "rank": 27, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "aisle", "segment": "general_english", "domain": null, "takes": null, "p_fail": 25, "pattern": "snap_to_neighbour", "likely_heard_as": ["isle", "aisles"], "reason": "Homophone confusion with isle.", "origin": "proposed", "rank": 28, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "jewelry", "segment": "general_english", "domain": null, "takes": null, "p_fail": 25, "pattern": "orthographic", "likely_heard_as": ["jewellery", "jewelery"], "reason": "Regional spelling variant.", "origin": "proposed", "rank": 29, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "forehead", "segment": "general_english", "domain": null, "takes": null, "p_fail": 25, "pattern": "word_boundary", "likely_heard_as": ["for head", "forehead's"], "reason": "Compound split into two words.", "origin": "proposed", "rank": 30, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "queue", "segment": "general_english", "domain": null, "takes": null, "p_fail": 25, "pattern": "orthographic", "likely_heard_as": ["cue", "Q"], "reason": "Homophone with cue and letter Q.", "origin": "proposed", "rank": 31, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "colonel", "segment": "general_english", "domain": null, "takes": null, "p_fail": 25, "pattern": "snap_to_neighbour", "likely_heard_as": ["kernel", "colonial"], "reason": "Unexpected pronunciation with near neighbour kernel.", "origin": "proposed", "rank": 32, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "buoy", "segment": "general_english", "domain": null, "takes": null, "p_fail": 25, "pattern": "snap_to_neighbour", "likely_heard_as": ["boy", "bui"], "reason": "Homophone with boy.", "origin": "proposed", "rank": 33, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "apparatus", "segment": "general_english", "domain": null, "takes": null, "p_fail": 25, "pattern": "snap_to_neighbour", "likely_heard_as": ["apparati", "a paratus"], "reason": "Word boundary and stress issue.", "origin": "proposed", "rank": 34, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "sovereign", "segment": "general_english", "domain": null, "takes": null, "p_fail": 25, "pattern": "snap_to_neighbour", "likely_heard_as": ["sovereignty", "sovran"], "reason": "Vowel reduction variants.", "origin": "proposed", "rank": 35, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "tangible", "segment": "general_english", "domain": null, "takes": null, "p_fail": 25, "pattern": "snap_to_neighbour", "likely_heard_as": ["tangibal", "tangle"], "reason": "Vowel reduction.", "origin": "proposed", "rank": 36, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "ingredient", "segment": "general_english", "domain": null, "takes": null, "p_fail": 20, "pattern": "snap_to_neighbour", "likely_heard_as": ["ingredients", "engredient"], "reason": "Initial vowel and plural ambiguity.", "origin": "proposed", "rank": 37, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "penguin", "segment": "general_english", "domain": null, "takes": null, "p_fail": 20, "pattern": "consonant_shift", "likely_heard_as": ["pengwin", "pegin"], "reason": "Unusual consonant cluster.", "origin": "proposed", "rank": 38, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "espresso", "segment": "general_english", "domain": null, "takes": null, "p_fail": 20, "pattern": "orthographic", "likely_heard_as": ["expresso", "express o"], "reason": "Common misspelling pattern with x and s.", "origin": "proposed", "rank": 39, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "ballpark", "segment": "general_english", "domain": null, "takes": null, "p_fail": 20, "pattern": "word_boundary", "likely_heard_as": ["ball park", "ballpark figure"], "reason": "Compound split.", "origin": "proposed", "rank": 40, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "Wednesday", "segment": "general_english", "domain": null, "takes": null, "p_fail": 20, "pattern": "snap_to_neighbour", "likely_heard_as": ["Wensday", "Wednesdays"], "reason": "Commonly reduced pronunciation.", "origin": "proposed", "rank": 41, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "February", "segment": "general_english", "domain": null, "takes": null, "p_fail": 20, "pattern": "snap_to_neighbour", "likely_heard_as": ["Febuary", "Febrary"], "reason": "Reduced medial syllable.", "origin": "proposed", "rank": 42, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "mischief", "segment": "general_english", "domain": null, "takes": null, "p_fail": 20, "pattern": "snap_to_neighbour", "likely_heard_as": ["mis chief", "mischiefs"], "reason": "Word boundary split risk.", "origin": "proposed", "rank": 43, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "debt", "segment": "general_english", "domain": null, "takes": null, "p_fail": 20, "pattern": "initial_drop", "likely_heard_as": ["det", "dept"], "reason": "Silent b drop.", "origin": "proposed", "rank": 44, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "sheriff", "segment": "general_english", "domain": null, "takes": null, "p_fail": 20, "pattern": "snap_to_neighbour", "likely_heard_as": ["sheriffs", "sherif"], "reason": "Spelling variant risk.", "origin": "proposed", "rank": 45, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "wrestle", "segment": "general_english", "domain": null, "takes": null, "p_fail": 20, "pattern": "initial_drop", "likely_heard_as": ["restle", "rustle"], "reason": "Silent w drop.", "origin": "proposed", "rank": 46, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "subtly", "segment": "general_english", "domain": null, "takes": null, "p_fail": 20, "pattern": "initial_drop", "likely_heard_as": ["subtlety", "sutely"], "reason": "Silent b and derivational confusion.", "origin": "proposed", "rank": 47, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "coup", "segment": "general_english", "domain": null, "takes": null, "p_fail": 20, "pattern": "snap_to_neighbour", "likely_heard_as": ["coop", "coupe"], "reason": "Vowel homophone confusion.", "origin": "proposed", "rank": 48, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "iceberg", "segment": "general_english", "domain": null, "takes": null, "p_fail": 15, "pattern": "word_boundary", "likely_heard_as": ["ice berg", "icebergs"], "reason": "Compound split into two words.", "origin": "proposed", "rank": 49, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "vegetable", "segment": "general_english", "domain": null, "takes": null, "p_fail": 15, "pattern": "snap_to_neighbour", "likely_heard_as": ["vegetables", "veggie table"], "reason": "Multisyllabic with reduced vowels.", "origin": "proposed", "rank": 50, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "fragile", "segment": "general_english", "domain": null, "takes": null, "p_fail": 15, "pattern": "snap_to_neighbour", "likely_heard_as": ["fragil", "fragility"], "reason": "Derivational variants.", "origin": "proposed", "rank": 51, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "conscious", "segment": "general_english", "domain": null, "takes": null, "p_fail": 15, "pattern": "snap_to_neighbour", "likely_heard_as": ["conscience", "conscious of"], "reason": "Homophone-like neighbour.", "origin": "proposed", "rank": 52, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "quarantine", "segment": "general_english", "domain": null, "takes": null, "p_fail": 15, "pattern": "snap_to_neighbour", "likely_heard_as": ["quarantined", "carantine"], "reason": "Initial consonant variation.", "origin": "proposed", "rank": 53, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "sword", "segment": "general_english", "domain": null, "takes": null, "p_fail": 15, "pattern": "initial_drop", "likely_heard_as": ["sord", "soared"], "reason": "Silent w drop.", "origin": "proposed", "rank": 54, "run": "haiku-r1", "model": "claude-haiku-5-5"}
+{"term": "dependent", "segment": "general_english", "domain": null, "takes": null, "p_fail": 10, "pattern": "none", "likely_heard_as": ["dependant"], "reason": "Spelling variant only.", "origin": "proposed", "rank": 55, "run": "haiku-r1", "model": "claude-haiku-5-5"}