Harshkmr/omniscribe_corpus

OmniScribe Corpus

A multilingual speech transcription corpus designed for fine-tuning ASR models on Indian medical and general-domain speech. It covers Hindi, Marathi, and Indian English, with a focus on clinical and healthcare contexts.

Overview

Split	Rows (after oversampling)	Approx. Duration
train	~30750	~230 hrs
benchmark	~4,089	~25 hrs

Audio samples average 20–30 seconds each. All samples are at least 5 seconds long.

Data Sources

Train Split

Source	Language	Unique Rows	Oversample	Effective Rows	Domain
ai4bharat/IndicVoices (hindi/train)	Hindi	9,000	×1	9,000	General
ai4bharat/IndicVoices (marathi/train)	Marathi	9,000	×1	9,000	General
OUTCOMESAI/medical_n_common_speech_corpus_50_50 (en/train)	English	9,000	×1	9,000	Medical
ekacare/eka-medical-asr-evaluation-dataset (en/test, rows 0–1499)	English	1,500	×2	3,000	Medical
ekacare/eka-medical-asr-evaluation-dataset (hi/test, rows 0–149)	Hindi	150	×5	750	Medical

Oversampling note: Ekacare rows are repeated (×2 EN, ×5 HI) in the train split to compensate for the small pool size and to up-weight high-quality clinical speech. Oversampled rows are spread evenly via a global shuffle (seed=42).

Benchmark Split

No oversampling is applied. Sources are held-out partitions not seen during training. | Source | Language | Rows | Domain | |--------|----------|------|--------| | ai4bharat/IndicVoices (hindi/valid) | Hindi | 600 | General | | ai4bharat/IndicVoices (marathi/valid) | Marathi | 600 | General | | OUTCOMESAI/medical_n_common_speech_corpus_50_50 (en/test) | English | 600 | Medical | | ekacare/eka-medical-asr-evaluation-dataset (en/test, rows 1500–3619) | English | 2,119 | Medical | | ekacare/eka-medical-asr-evaluation-dataset (hi/test, rows 150–319) | Hindi | 170 | Medical |

Schema

Each row contains the following columns: | Column | Type | Description | |--------|------|-------------| | audio | Audio | Audio sample (sampling rate varies by source) | | text | string | Ground-truth transcription | | language | string | One of hindi, marathi, english | | source | string | Source dataset identifier (e.g. IndicVoices-train, ekacare-test-en) | | duration | float | Audio duration in seconds | | system_prompt | string | Role-specific instruction prompt for the transcription model |

System Prompts

Each row carries a system_prompt tailored to its source and language. These are designed for instruction-tuned models (e.g. Whisper, Gemini, Qwen-Audio) that accept a system/instruction context alongside the audio. | Source key | Language | Persona | |------------|----------|---------| | IndicVoices-train | Hindi | General clerk, Uttar Pradesh | | IndicVoices-train | Marathi | General clerk, Maharashtra | | OUTCOMESAI-train | English | Medical clerk, Delhi | | ekacare-test-en | English | Medical clerk, hospital in Delhi | | ekacare-test-hi | Hindi | Medical clerk, hospital in Delhi (Hindi + English medical terms allowed) | All prompts instruct the model to use [INAUDIBLE] for uncertain words and to return only the transcription with no additional commentary.

Construction

The corpus was built using a two-stage pipeline: 1. build_corpus — streams each source dataset independently, filters samples shorter than 5 seconds, shuffles with a 200-sample buffer (seed=42), and uploads to five separate per-language HF repos in 300-row Parquet chunks. Each repo can be rebuilt independently on failure. 2. concat_corpus — loads all per-language repos, applies oversampling to the train split, concatenates, shuffles globally (seed=42), and pushes to this final repo. Per-language intermediate repos: - Harshkmr/omniscribe_indic_hindi - Harshkmr/omniscribe_indic_marathi - Harshkmr/omniscribe_outcomesai - Harshkmr/omniscribe_ekacare_en - Harshkmr/omniscribe_ekacare_hi

Usage

from datasets import load_dataset
 
# Full train split
ds_train = load_dataset("Harshkmr/omniscribe_corpus", split="train")
 
# Benchmark split
ds_bench = load_dataset("Harshkmr/omniscribe_corpus", split="benchmark")
 
# Filter by language
hindi_train = ds_train.filter(lambda x: x["language"] == "hindi")

Using with a model

# The system_prompt field is ready to pass directly to instruction-tuned ASR models
sample = ds_train[0]
print(sample["system_prompt"])
# → "You are a helpful general clerk from India who is fluent in Hindi ..."
 
audio  = sample["audio"]   # {"array": np.ndarray, "sampling_rate": int}
target = sample["text"]    # ground-truth transcription

---

Intended Use

• Fine-tuning multilingual ASR models on Indian-accented speech
• Benchmarking transcription quality on medical vs. general-domain audio
• Instruction-tuning audio-language models with role-specific prompts

---

Limitations

• Ekacare rows are repeated in the train split (×2 EN, ×5 HI). Models may memorise these samples; exclude the ekacare-* source rows if you need a fully deduplicated train set.
• No Bhojpuri in the current build (infrastructure is in place; data sourcing is pending).
• Audio quality varies across sources — IndicVoices is crowdsourced field audio; OUTCOMESAI and Ekacare are semi-controlled clinical recordings.
• This corpus is private and intended for internal model development only. Downstream licensing must comply with each upstream dataset's terms.

---

Citation

If you use this corpus, please cite the upstream datasets:

@misc{omniscribe_corpus,
  title  = {OmniScribe Corpus},
  author = {Harshkmr},
  year   = {2025},
  note   = {Compiled from ai4bharat/IndicVoices, OUTCOMESAI/medical\_n\_common\_speech\_corpus\_50\_50, and ekacare/eka-medical-asr-evaluation-dataset}
}