metadata

README.md

@@ -25,11 +25,11 @@ configs:
   - config_name: default
     data_files:
       - split: train
-        path: metadata.csv
+        path: data/**
   - config_name: original
     data_files:
       - split: train
-        path: metadata_original.csv
+        path: data_original/**
 dataset_info:
   - config_name: default
     features:
@@ -45,14 +45,10 @@ dataset_info:
         dtype: string
       - name: gender
         dtype: string
-      - name: age
-        dtype: int64
       - name: pipeline
         dtype: string
       - name: sentence_id
         dtype: int64
-      - name: duration_s
-        dtype: float64
   - config_name: original
     features:
       - name: audio
@@ -67,18 +63,10 @@ dataset_info:
         dtype: string
       - name: gender
         dtype: string
-      - name: age
-        dtype: int64
       - name: pipeline
         dtype: string
       - name: sentence_id
         dtype: int64
-      - name: duration_s
-        dtype: float64
-      - name: sample_rate
-        dtype: int64
-      - name: channels
-        dtype: int64
 ---
 
 - **Developed by:**Researcher at Dept. of CSE, SUST, Bangladesh
@@ -91,14 +79,14 @@ dataset_info:
 > This dataset is released ahead of its companion manuscript,
 > *"BPAC: How Detectable Is Regional Accent in Read Bangla? A Parallel
 > Corpus and Its Phonetic Structure."* Please cite the **dataset directly**
-> (see [Citation](#citation) below) until the paper is published — this
+> (see [Citation](#citation) below) until the paper is published, this
 > card will be updated with the full citation at that point.
 
 ## Dataset summary
 
 BPAC is a **parallel**, regional-accent speech corpus for Bangla (Bengali):
-**51 native speakers from 8 regions of Bangladesh** — Sylhet, Chattogram,
-Dhaka, Faridpur, Khulna, Mymensingh, Rajshahi, and Rangpur — each read an
+**51 native speakers from 8 regions of Bangladesh**; Sylhet, Chattogram,
+Dhaka, Faridpur, Khulna, Mymensingh, Rajshahi, and Rangpur; each read an
 **identical, phonetically diverse 250-sentence script**, so that region and
 speaker vary while linguistic content is held fixed by design. The result
 is **12,747 utterances** (three recordings were excluded as unusably short
@@ -110,7 +98,7 @@ Bangla speech corpus we are aware of: it is the only one, to our knowledge,
 in which the *same* text is read in full by a defined sample of speakers
 across multiple named regions, making it usable for accent detectability
 research, accent/voice conversion (a same-content reference exists for
-every utterance), and accent-controlled ASR evaluation — not only
+every utterance), and accent-controlled ASR evaluation, not only
 general-purpose ASR acoustic modeling.
 
 - **Speakers**: 51 (25 female, 26 male), aged 19–25
@@ -172,45 +160,68 @@ root folder and the underlying audio differ.
 
 
 
+## Composition by region
+ 
+Every speaker read all 250 sentences. There are no missing
+(speaker, sentence) cells beyond the three excluded empty files.
+ 
+| Region | Male | Female | Speakers | Utterances | Audio (min) | Mean utt. (s) |
+|---|---|---|---|---|---|---|
+| Sylhet | 6 | 4 | 10 | 2,500 | 297.2 | 7.13 |
+| Faridpur | 4 | 3 | 7 | 1,749 | 231.1 | 7.92 |
+| Dhaka | 3 | 3 | 6 | 1,500 | 210.2 | 8.41 |
+| Chattogram | 3 | 3 | 6 | 1,500 | 206.7 | 8.27 |
+| Khulna | 3 | 3 | 6 | 1,500 | 200.3 | 8.01 |
+| Rajshahi | 3 | 3 | 6 | 1,500 | 190.3 | 7.61 |
+| Mymensingh | 2 | 3 | 5 | 1,248 | 167.0 | 8.02 |
+| Rangpur | 2 | 3 | 5 | 1,250 | 161.6 | 7.76 |
+| **Total** | **26** | **25** | **51** | **12,747** | **1,664.5** | **7.83** |
+ 
+
+
+
 **Important confound to be aware of**: recording pipeline is perfectly
-confounded with gender in this corpus — every lab recording is male, every
+confounded with gender in this corpus. Every lab recording is male, every
 app recording is female. No claim about pipeline/channel effects can be
 separated from gender effects using this corpus alone; any apparent
 "pipeline effect" is equally validly read as a "gender effect," and vice
 versa.
 
-## How to load
 
+
+## How to load
+ 
 ```python
 from datasets import load_dataset
-
+ 
+# 16 kHz working corpus (default)
 ds = load_dataset("ArfatulAsif/BPAC")
+ 
+# original-rate recordings
+ds_orig = load_dataset("ArfatulAsif/BPAC", "original")
+ 
 print(ds["train"][0])
 ```
 
-Because every file is annotated in `metadata.csv`, the `datasets`
-`audiofolder` loader picks up `region`, `gender`, `pipeline`, `sentence_id`,
-and the `sentence` transcript automatically alongside the **16 kHz**
-`file_name` audio — no custom loading script needed. `file_name_original`
-is carried through as a plain string column (`datasets` does not
-auto-decode a second audio column from `audiofolder`); load the
-original-rate audio explicitly when you need it:
-
+Both configs decode audio automatically, so `ds["train"][0]["audio"]`
+returns a dict containing the decoded waveform array and its sampling
+rate, alongside every metadata field.
+ 
+To retrieve every speaker's recording of a specific sentence the core
+parallel-corpus operation:
+ 
 ```python
-import soundfile as sf
-
-row = ds["train"][0]
-orig_audio, orig_sr = sf.read(row["file_name_original"])
-# or, if you downloaded the repo locally rather than via load_dataset:
-# orig_audio, orig_sr = sf.read(os.path.join(local_repo_dir, "data_original", ...))
+sentence_42 = ds["train"].filter(lambda x: x["sentence_id"] == 42)
 ```
-
-To retrieve every region's recording of a specific sentence (the core
-parallel-corpus operation):
-
+ 
+To work within a single region or recording pipeline:
+ 
 ```python
-sentence_42 = ds["train"].filter(lambda x: x["sentence_id"] == 42)
+sylhet = ds["train"].filter(lambda x: x["region"] == "sylhet")
+lab_only = ds["train"].filter(lambda x: x["pipeline"] == "lab")
 ```
+ 
+
 
 ## Audio format
 

metadata.csv → data/metadata.csv

@@ -1,4 +1,4 @@
-audio,utterance_id,speaker_id,region,gender,pipeline,sentence_id,sentence
+file_name,utterance_id,speaker_id,region,gender,pipeline,sentence_id,sentence
 data/chattogram/chattogram_female_001_001.wav,chattogram_female_001_001,chattogram_female_001,chattogram,female,app,1,গত শুক্রবার তাঁরা প্রায় এক ঘণ্টা ধরে হোয়াইট হাউসে বৈঠক করেছেন।
 data/chattogram/chattogram_female_001_002.wav,chattogram_female_001_002,chattogram_female_001,chattogram,female,app,2,ভারতের হয়ে একাই চার উইকেট শিকার করেছেন কুলদীপ যাদব।
 data/chattogram/chattogram_female_001_003.wav,chattogram_female_001_003,chattogram_female_001,chattogram,female,app,3,শুরুতে শুধু ফোনে ভিডিও করতেন।

metadata_original.csv → data_original/metadata.csv

@@ -1,4 +1,4 @@
-audio,utterance_id,speaker_id,region,gender,pipeline,sentence_id,sentence
+file_name,utterance_id,speaker_id,region,gender,pipeline,sentence_id,sentence
 data_original/chattogram/chattogram_female_001_001.wav,chattogram_female_001_001,chattogram_female_001,chattogram,female,app,1,গত শুক্রবার তাঁরা প্রায় এক ঘণ্টা ধরে হোয়াইট হাউসে বৈঠক করেছেন।
 data_original/chattogram/chattogram_female_001_002.wav,chattogram_female_001_002,chattogram_female_001,chattogram,female,app,2,ভারতের হয়ে একাই চার উইকেট শিকার করেছেন কুলদীপ যাদব।
 data_original/chattogram/chattogram_female_001_003.wav,chattogram_female_001_003,chattogram_female_001,chattogram,female,app,3,শুরুতে শুধু ফোনে ভিডিও করতেন।