danielrosehill's picture
One dataset, two tabs: sentences with take counts, and takes
8b49c3b verified
|
Raw History Blame Contribute Delete
5.63 kB
---
license: cc-by-4.0
language:
- en
task_categories:
- automatic-speech-recognition
pretty_name: Phone Number Sentences
tags:
- speech
- asr
- fine-tuning
- single-speaker
- phone-numbers
- english
size_categories:
- n<1K
configs:
- config_name: takes
default: true
data_files:
- split: train
path: data/train-*.parquet
- split: test
path: data/test-*.parquet
- config_name: sentences
data_files:
- split: train
path: sentences/train.parquet
- split: test
path: sentences/test.parquet
---
# Phone Number Sentences
<!-- dataset-card-frame:begin -->
[![Type: Audio](https://img.shields.io/badge/Type-Audio-1f6feb)](https://huggingface.co/datasets/Hebrish/Dataset-Index) [![Role: Split-out](https://img.shields.io/badge/Role-Split--out-6e7781)](https://huggingface.co/datasets/Hebrish/Dataset-Index) [![Subset of: Read-Aloud-Voice-Sentences](https://img.shields.io/badge/Subset_of-Read--Aloud--Voice--Sentences-0e8a16)](https://huggingface.co/datasets/danielrosehill/Read-Aloud-Voice-Sentences)
**Audio dataset, split-out.** Rows copied from the master, [`danielrosehill/Read-Aloud-Voice-Sentences`](https://huggingface.co/datasets/danielrosehill/Read-Aloud-Voice-Sentences), so one task can be browsed alone. Training reads the master, not this copy. Every audio dataset as one tree, with hours: [the audio index](https://huggingface.co/datasets/danielrosehill/Audio-Dataset-Index). How every dataset fits together: [the dataset index](https://huggingface.co/datasets/Hebrish/Dataset-Index).
<!-- dataset-card-frame:end -->
One speaker dictating **phone numbers** inside everyday sentences, read digit by digit (oh or zero), with double and triple, or in pairs. Takes up to capture app release 236 follow a script, kept as `spoken_text`; later takes are read the speaker's own way from the written number (`reading` natural, no `spoken_text`). `text` writes the number grouped with hyphens: Israeli mobile `054-332-1879`, landline `02-624-3318`, `1-800-738-702`, Irish `087-412-6630` / `01-399-5725`, UK `07798-388821`, US `718-206-5332`, `extension 802`. International rows say "plus": `text` is `plus 972 50-913-6053` (Parakeet v2 has no +) and `text_full` is `+972 50-913-6053`. Control sentences (`prompt_kind` control) use double, triple and oh in their ordinary sense (a double espresso, triple-check). Prompts self-written by Claude (2026-10-28).
**Website: [hebrish.app](https://hebrish.app)**: get the Hebrish keyboard for Android and the desktop apps, see the results, and find every model, dataset and demo in one place.
**Two tabs.** `sentences` lists every ground-truth sentence with a `takes` count, so
`takes = 0` shows what has not been recorded. `takes` has one row per recording with its audio, joined on `sentence_id`; a
sentence can have several takes. `load_dataset("Hebrish/Phone-Number-Sentences", "sentences")` reads the text; the default config reads the takes.
Takes come from [`danielrosehill/Read-Aloud-Voice-Sentences`](https://huggingface.co/datasets/danielrosehill/Read-Aloud-Voice-Sentences), the full dataset, with the same ids and
train/test split, by `scripts/sync_subsets.py` in the project repository. Do not edit this dataset directly. New takes append and
only the sentence counts are rewritten.
**Speaker:** Daniel Rosehill, the dataset's author, who was born in Ireland and speaks English with an **Irish accent**, and who lives in Jerusalem. Every recording is his voice. **Deliberate variation:** he sometimes changes his speaking speed on purpose (fast and rushed, or slow) and sometimes says a word the way people commonly mispronounce it, to make the models robust to real speech. The transcript is always the sentence as written, so a take whose speed or pronunciation sounds off against its card is usually intentional, not a recording error.
All audio was recorded on a **OnePlus Nord 3 5G** phone using its **internal microphone**, with browser echo cancellation, noise suppression and auto gain turned off.
## Layout
```
sentences/train.parquet one row per sentence: `id`, `text`, `takes`, `prompt_ids`, `served` and the sentence columns
data/train-*.parquet one row per take: `audio` (the FLAC or WAV file, embedded) and every column
meta/train.jsonl index: the same rows without audio, plus `shard`, `audio_path` and the audio's byte range
meta/schema.json column types
data/test-*.parquet, meta/test.jsonl held-out set, never trained on
```
Audio is 48 kHz mono 24-bit FLAC for rows from 2026-10-22 and 16 kHz mono 16-bit WAV before.
The `test` split is **term-disjoint**: a held-out term never appears in `train`.
`load_dataset("Hebrish/Phone-Number-Sentences")` works directly.
## Columns
A take's `sentence_id` is its row's `id` in `sentences`. A sentence with `served: false` is no longer offered by the app; it is kept because a take used it. The columns are the same as in the main dataset: `text` (ground truth, punctuated and cased like
Parakeet v2 output), `text_he` (the same transcript with its Hebrew terms in Hebrew script, for
Hebrew-capable models), `categories`, `target_terms`, `prompt_source`, `prompt_origin` (manual / dataset / generated), `edited`, `duration_s`,
`environment`, `conditions`, `device`, `speaker`, `speaker_gender`, `contributor`, `hebrew_pronunciation`, `variant_groups`, `variant_spelling` and `recorded_at`. The main dataset's card describes them.
## Licence
CC-BY-4.0, with sentence-text attribution per `prompt_source` as described in the
[main dataset's licence section](https://huggingface.co/datasets/danielrosehill/Read-Aloud-Voice-Sentences#licence-and-attribution).