Update README.md

README.md

@@ -2,11 +2,11 @@
 language:
 - hi
 license: cc0-1.0
-pretty_name: Hindi ASR 1k (Common Voice subset)
+pretty_name: Hindi ASR 11k (Common Voice subset)
 task_categories:
 - automatic-speech-recognition
 size_categories:
-- 1K<n<10K
+- 10K<n<100K
 tags:
 - hindi
 - indic
@@ -14,7 +14,7 @@ tags:
 - speech
 - common-voice
 - whisper
-- groundtruth
+- finetune
 annotations_creators:
 - crowdsourced
 language_details: hi-IN
@@ -22,74 +22,66 @@ source_datasets:
 - mozilla-foundation/common_voice
 ---
 
-# Hindi ASR 1k
+# Hindi ASR 11k
 
-A ready-to-use **Hindi automatic speech recognition (ASR)** dataset: ~10,000
-short read-speech utterances with ground-truth Devanagari transcriptions, derived
-from **[Mozilla Common Voice](https://commonvoice.mozilla.org) (Hindi)**. It is
-sized for quick fine-tuning experiments and for benchmarking/word-error-rate (WER)
-evaluation of models such as [Whisper](https://github.com/openai/whisper) and
-other multilingual/Indic ASR systems — small enough to iterate on a single GPU,
-while still being a real, human-spoken corpus. The `train/audio/` and
-`test/audio/` folders additionally provide a curated **1,000-clip** subset under
-sequential `cv_*` IDs (the "1k" in the name).
+A **Hindi automatic speech recognition (ASR)** dataset of ~10,000 short read-speech
+utterances with ground-truth Devanagari transcriptions, derived from
+**[Mozilla Common Voice](https://commonvoice.mozilla.org) (Hindi)**. It is a
+convenient, dependency-light corpus for **fine-tuning** and **WER benchmarking** of
+[Whisper](https://github.com/openai/whisper) and other multilingual / Indic ASR
+models — large enough to move the needle when fine-tuning, small enough to iterate
+on a single GPU.
 
 - **Language:** Hindi (`hi`), Devanagari script
 - **Task:** Automatic Speech Recognition (speech → text)
 - **Audio:** MP3, mono
-- **Splits:** `train` (9,000 transcribed clips) / `test` (1,000 transcribed clips)
+- **Splits:** `train` (9,000 clips) / `test` (1,000 clips)
 - **Source:** Mozilla Common Voice Hindi (crowd-sourced read speech)
 - **License:** CC0-1.0 (public domain), inherited from Common Voice
 
 ## Dataset structure
 
-Each split ships audio under `clips/` (original Common Voice filenames) with a
-metadata file mapping every clip to its transcript. A parallel `audio/` folder
-holds the same kind of clips under sequential `cv_NNNNNN.mp3` identifiers.
+Each split contains an audio folder and a metadata file that maps every clip to its
+transcription. Every metadata row has a matching audio file.
 
 ```
-hi-asr-1k/
+11k-asr/
 ├── train/
-│   ├── clips/           # 9,000 common_voice_hi_*.mp3
-│   ├── audio/           #   800 cv_*.mp3  (sequential IDs)
-│   └── metadata.tsv     # file_name, text  (transcripts for the clips/)
+│   ├── clips/         # 9,000 common_voice_hi_*.mp3
+│   └── metadata.tsv   # file_name, text
 └── test/
-    ├── clips/           # 1,000 common_voice_hi_*.mp3
-    ├── audio/           #   200 cv_*.mp3  (sequential IDs)
-    ├── metadata.csv     # file_name, text  (transcripts for the clips/)
-    └── metadata.tsv     # path, text       (transcripts for the audio/ files)
+    ├── clips/         # 1,000 common_voice_hi_*.mp3
+    └── metadata.tsv   # file_name, text
 ```
 
 ### Data fields
 
-- **`file_name`** / **`path`** (`string`): relative path to the audio clip.
+- **`file_name`** (`string`): audio filename, relative to the split's `clips/` folder.
 - **`text`** (`string`): the ground-truth Hindi transcription in Devanagari.
-- **`audio`** (`Audio`): the decoded waveform + sampling rate, when loaded via the
-  🤗 `datasets` `AudioFolder` builder.
+- **`audio`** (`Audio`): decoded waveform + sampling rate, once loaded (see below).
 
 ### Data instance
 
 ```json
 {
-  "file_name": "common_voice_hi_24663092.mp3",
-  "text": "उन्हे अपने अकलमंद बेटे पर नाज़ है।"
+  "file_name": "common_voice_hi_27545949.mp3",
+  "text": "जिम मेरे घर में ठहरा था।"
 }
 ```
 
 ### Splits
 
-| Split | Transcribed clips (`clips/`) | Extra audio (`audio/`) |
-| ----- | ---------------------------: | ---------------------: |
-| train | 9,000                        | 800                    |
-| test  | 1,000                        | 200                    |
-
-Every row in each split's metadata has a matching audio file in `clips/`.
+| Split | Clips |
+| ----- | ----: |
+| train | 9,000 |
+| test  | 1,000 |
+| **total** | **10,000** |
 
 ## Usage
 
-The metadata is tab-separated and its `file_name` column is relative to the
+The metadata is **tab-separated** and its `file_name` column is relative to the
 `clips/` folder, so the most reliable way to load audio + text together is to
-download the repo and join the paths yourself:
+download the repo and join the paths:
 
 ```python
 import os
@@ -97,18 +89,20 @@ import pandas as pd
 from huggingface_hub import snapshot_download
 from datasets import Dataset, Audio
 
-root = snapshot_download("dhruvkys/hi-asr-1k", repo_type="dataset")
+root = snapshot_download("dhruvkys/11k-asr", repo_type="dataset")
 
-split = "train"                                  # or "test"
-meta = "metadata.tsv" if split == "train" else "metadata.csv"  # both are TSV-formatted
-df = pd.read_csv(os.path.join(root, split, meta), sep="\t")
-df["audio"] = df["file_name"].apply(
-    lambda f: os.path.join(root, split, "clips", f)
-)
+def load_split(split):
+    df = pd.read_csv(os.path.join(root, split, "metadata.tsv"), sep="\t")
+    df["audio"] = df["file_name"].apply(
+        lambda f: os.path.join(root, split, "clips", f)
+    )
+    return Dataset.from_pandas(df).cast_column("audio", Audio(sampling_rate=16_000))
+
+train = load_split("train")
+test  = load_split("test")
 
-ds = Dataset.from_pandas(df).cast_column("audio", Audio(sampling_rate=16_000))
-print(ds[0]["text"])
-print(ds[0]["audio"]["array"].shape, ds[0]["audio"]["sampling_rate"])
+print(train[0]["text"])
+print(train[0]["audio"]["array"].shape, train[0]["audio"]["sampling_rate"])
 ```
 
 Just need the transcripts? Read the metadata directly:
@@ -117,7 +111,7 @@ Just need the transcripts? Read the metadata directly:
 import pandas as pd
 
 df = pd.read_csv(
-    "hf://datasets/dhruvkys/hi-asr-1k/test/metadata.csv",
+    "hf://datasets/dhruvkys/11k-asr/train/metadata.tsv",
     sep="\t",                     # metadata is tab-separated
 )
 print(df.head())
@@ -133,8 +127,8 @@ asr = pipeline("automatic-speech-recognition",
                model="openai/whisper-small", generate_kwargs={"language": "hi"})
 wer = evaluate.load("wer")
 
-preds = [asr(x["audio"])["text"] for x in ds]
-refs  = [x["text"] for x in ds]
+preds = [asr(x["audio"])["text"] for x in test]
+refs  = [x["text"] for x in test]
 print("WER:", wer.compute(predictions=preds, references=refs))
 ```
 
@@ -142,8 +136,8 @@ print("WER:", wer.compute(predictions=preds, references=refs))
 
 Audio and transcriptions originate from the **Mozilla Common Voice** Hindi
 corpus — sentences read aloud and contributed by volunteers, then validated by
-community review. This dataset is a curated ~1k-per-split subset repackaged for
-convenient fine-tuning and evaluation; no new recordings were made.
+community review. This dataset is a curated subset repackaged for convenient
+fine-tuning and evaluation; no new recordings were made.
 
 ## Preprocessing recommendations
 
@@ -179,12 +173,11 @@ If you use this dataset, please cite Mozilla Common Voice:
 
 ## Known limitations
 
-- **Small scale.** ~1k clips per split — good for prototyping and benchmarking, not
-  for training a production ASR model from scratch.
+- **Modest scale.** ~10k clips — strong for fine-tuning and benchmarking, but not a
+  substitute for the full Common Voice corpus when training from scratch.
 - **Domain skew.** Many sentences are news headlines, so vocabulary leans toward
   named entities, politics, and current-affairs terms.
-- **Two naming schemes.** `clips/` uses original Common Voice filenames; `audio/`
-  uses sequential `cv_*` IDs. The `train/audio/` clips do not ship a dedicated
-  transcript file (only `test/audio/` has `test/metadata.tsv`).
-- **Metadata format.** The metadata files are **tab-separated**, even the one named
-  `metadata.csv`. Pass `sep="\t"` when reading them (see usage above).
+- **Metadata format.** Metadata files are **tab-separated** (`.tsv`); pass
+  `sep="\t"` when reading them. Because `file_name` is not prefixed with `clips/`,
+  the plain `audiofolder` builder will not auto-attach transcripts — use the
+  snapshot-based loader above.