Host audio in the repo, drop the S3 URLs, remove recordings with PII

The Filename column held URLs into a live production storage bucket, and each URL carried the internal message id of the farmer's chat message. That id joined 5,346 of these rows to real conversations, users and districts, so publishing it gave anyone with database access a route from a public benchmark row back to an identified farmer. Audio now ships from this repo as audio/00001.ogg upward and Filename holds that path. The column name is unchanged, so existing code keeps working.

Every transcript was screened for personal information by two models, the second reviewing every flag plus a blind sample of the clears. 56 recordings were removed. A spoken name cannot be cropped out the way a watermark can, so removal was the only option.

10864 rows -> 10808 rows, 10808 audio files.

README.md

@@ -1,30 +1,101 @@
 ---
 license: cc-by-4.0
 task_categories:
-- audio-classification
-- automatic-speech-recognition
+  - automatic-speech-recognition
+  - audio-classification
 language:
-- hi
-- te
-- or
+  - hi
+  - te
+  - or
 size_categories:
-- 10K<n<100K
+  - 10K<n<100K
+configs:
+  - config_name: default
+    data_files:
+      - split: train
+        path: benchmark.parquet
 ---
 
+# Agri STT Benchmarking Dataset
 
-This is a domain-specific, multilingual agricultural speech dataset with a primary focus on Hindi, Telugu, and Odia, designed for speech-to-text and automatic speech recognition (ASR) tasks. It features human-annotated transcriptions and is intended for benchmarking ASR model performance in real-world agricultural scenarios.
+10,808 farmer voice queries in Hindi, Telugu and Odia, with reference transcripts, for benchmarking automatic speech recognition in agricultural contexts. The audio is included in this repository.
 
-This paper presents a comprehensive benchmark of 10 ASR models for agricultural advisory use across Hindi, Telugu, and Odia, using 10,934 real-world Farmer.Chat audio recordings with human-annotated transcripts. It introduces Agriculture Weighted Word Error Rate (AWWER) and LLM-based utility scoring to better evaluate domain-critical agricultural terminology beyond traditional WER, CER, and MER metrics.
+Every recording is a smallholder farmer speaking a question to [Farmer.Chat](https://farmerchat.digitalgreen.org/), an AI advisory service run by [Digital Green](https://www.digitalgreen.org/). Reference transcripts were produced by human annotators. Nothing here is read from a script or recorded in a studio, so the audio carries the noise, overlap and dialect of real field conditions.
 
-Publishing Context:
-The research paper titled “Benchmarking Automatic Speech Recognition for Indian Languages in Agricultural Contexts” is published on arXiv (arXiv:2602.03868), and the accompanying agricultural ASR benchmark dataset is publicly released on Hugging Face to support reproducible research and community-driven development.
+## Contents
 
-## Reference
+- `audio/` - 10,808 OGG files, numbered `00001.ogg` upward and split across subdirectories of 1,000 (`audio/000/`, `audio/001/`, and so on). Use the `Filename` column rather than assembling the path yourself.
+- `benchmark.parquet` - all three languages in one table
+- `Agri_Benchmark_Hindi.csv` - Hindi only, 4,576 rows
+- `Agri_Benchmark_Telugu.csv` - Telugu only, 4,027 rows
+- `Agri_Benchmark_Oriya.csv` - Odia only, 2,205 rows
 
-If you use this dataset, please cite:
+| Column | Description |
+|--------|-------------|
+| `Filename` | Path to the audio file within this repository |
+| `Language` | BCP-47 language code of the utterance |
+| `reference` | Human reference transcript |
 
-Paper link:
+## Splits
 
-https://arxiv.org/abs/2602.03868
+| Language | Code | Utterances |
+|---|---|---:|
+| Hindi | `hi` | 4,576 |
+| Telugu | `te` | 4,027 |
+| Odia | `or` | 2,205 |
+| **Total** | | **10,808** |
 
-https://arxiv.org/pdf/2602.03868
+## Usage
+
+```python
+from datasets import load_dataset
+from huggingface_hub import snapshot_download
+
+ds = load_dataset("DigiGreen/Agri_STT_Benchmarking_Dataset", split="train")
+root = snapshot_download("DigiGreen/Agri_STT_Benchmarking_Dataset", repo_type="dataset")
+
+row = ds[0]
+audio_path = f'{root}/{row["Filename"]}'
+print(row["Language"], row["reference"])
+```
+
+## Privacy
+
+Earlier versions of this dataset delivered the audio as links to a cloud storage bucket, and the filename in each link carried an internal message identifier. Both are gone. The audio now ships from this repository under sequential filenames that carry no information about the speaker, the conversation, or the original record. The `Filename` column keeps its name and now holds a path inside this repository.
+
+Every reference transcript was screened for personal information by two language models independently, the second reviewing everything the first flagged plus a blind sample of what it cleared. 56 recordings were removed because the speech contained personal information: a person's name, a village or landmark specific enough to narrow down a household, a phone or ID number, money tied to a named person, a medical detail, or a private domestic matter.
+
+| Reason for removal | Recordings |
+|---|---:|
+| place name | 31 |
+| person name | 16 |
+| health | 9 |
+| household | 6 |
+| money | 1 |
+| phone | 1 |
+
+Two limits are worth stating rather than leaving implied. The screening read the transcripts, not the audio, and a recording holds everything its transcript holds plus whatever the annotator left out. And a voice is biometric whatever the words are: these are real people, identifiable to anyone who knows them, which is the nature of a speech corpus rather than a defect in this one. Treat the recordings as personal data. If you find something that should not be here, please open a discussion and it will be removed.
+
+Some recordings also capture speech that is not the farmer's question, such as a second person in the room or a field agent explaining the app. Where that speech carried nothing identifying it was left in, because it is part of the real acoustic conditions this benchmark is meant to measure. Expect it when scoring against the references.
+
+## Changelog
+
+**July 2026.** Audio moved into the repository as `audio/00001.ogg` upward; the `Filename` column now holds that path instead of a storage URL. Added `benchmark.parquet` covering all three languages. Screened every transcript for personal information and removed 56 recordings.
+
+## Citation
+
+The accompanying paper, *Benchmarking Automatic Speech Recognition for Indian Languages in Agricultural Contexts*, is included in this repository ([arXiv:2602.03868](https://arxiv.org/abs/2602.03868)).
+
+```bibtex
+@dataset{digigreen_agri_stt_2026,
+  title={Agri STT Benchmarking Dataset},
+  author={Digital Green},
+  year={2026},
+  url={https://huggingface.co/datasets/DigiGreen/Agri_STT_Benchmarking_Dataset},
+  license={CC-BY-4.0}
+}
+```
+
+## License
+
+Released under [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/). You are free to share and adapt the data for any purpose, provided you give appropriate credit to Digital Green.

audio/000/00001.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8e545fdece09e44bc1c7b32f6f6669509c275e92434a746498f024d3c1772587
+size 12167

audio/000/00002.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:186ede72244e39a186760198484e98b38c31113540c9b7bd296e7a5c88f261cc
+size 31204

audio/000/00003.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ebbc631fd15ccfcd784c22aaa41cd82e00a6e194008ef0d9945db3ef2fde0715
+size 14706

audio/000/00004.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8dc8d51a85dee1b87109d1b3ac67a52dc577911c902a20c0c3d744956148c3d8
+size 15899

audio/000/00005.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b2229c85cf4c14886bf2075e1334d5f39dc36520d15515f3f8893eeda62376df
+size 21330

audio/000/00006.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c73d1aa831c1ca5ff8179f1a4769914f8c6325a1a50c066cfe59bdb597e06550
+size 16743

audio/000/00007.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e33c73b8dfe30ab5f51291f6b34dcf2ddb94233d75da11349cc12415a103ccc4
+size 32239

audio/000/00008.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d8f0e76a8a5bc509a39331c45fe8fc7344afd5d99dbc4918d830d99827a783fa
+size 15199

audio/000/00009.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3612b36e7777b782b06bf326f69b2d6949827a8a821ae420b4d28a46d76d4a86
+size 15958

audio/000/00010.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6ac3ad481f17941c657af8fe25204a4fe84a6d5ea1742703ca145ce8ef93b597
+size 18471

audio/000/00011.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8c21263d29fbd4e7deec38cc34bb52e4a4346bb6fed863ad54ad97de8bfbe68a
+size 11631

audio/000/00012.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d14a4e022f93ed889e8367057a9cf6503337d1ab85eb3695f5588fdc7ab6a114
+size 14607

audio/000/00013.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aeb50796c7264c532de1b258d0849f27bbb573ed221abdde8e7e341ee02aa0fe
+size 22342

audio/000/00014.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1383e7df47639a71e9d1656f0fa7628b1262a5b33f374f9addbc4ff350852cdc
+size 9302

audio/000/00015.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a0109834f71bf9f45e6962e118c93e97383d33351071f85b5dc5e4e0af517ac9
+size 13998

audio/000/00016.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:20b5f12a817e5dda8dab5aa5a619d7a74ddedd1087f48d4468a116a4573e6755
+size 30196

audio/000/00017.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58e8a210681d7c831c52613e64a68a43476f7e54bdf13c6ba43f458b45f1e061
+size 19949

audio/000/00018.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:554d67b791e27fbad9f3a33bb3628e8fc8c6f5070593f2bdf0d22ae34d6b1f93
+size 10248

audio/000/00019.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:566868e26fc45a07acd6127f4e8249521940cb6255d2ae81ac57260901a764be
+size 20074

audio/000/00020.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19f160a6a94cd88ecd7ccceaa4b745265c94a6613903ac48901dcd27cd3973dc
+size 18917

audio/000/00021.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:67e0785f214cb984cfd2b3478e4da421b9258499fcabf3ee8405b1fe291b9f36
+size 12629

audio/000/00022.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:868aed5302c64f74db27bc685ad392a54f5fdfabbe253a7f289246212ff44efb
+size 9378

audio/000/00023.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3babe011fcc8fa52a98a540f7b99dadbfb13702b7b990b9f416eb684bb578ae8
+size 17640

audio/000/00024.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ced218817ec2c22ceebac5d2cb2697656ed2b9c6af9c2d9b4877fc427497ec77
+size 11988

audio/000/00025.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b0f6ca2dd87121ad8534092dedd4d280efe8f601b85960deefe8eea3e99ef3fc
+size 12927

audio/000/00026.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d196d0f6861a117c734f281a893e8d34ee4947a683c1ebfdf81780cc0955b67d
+size 23473

audio/000/00027.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f69a3d0e68762a6e3a478bd1b244dc661fa8a97bd37b83c3fb0bb2b9f9511c08
+size 18751

audio/000/00028.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ec21fd0caa41088fda82a3c23f77f82d194eec3621e4c21db172c53eceda0de
+size 19978

audio/000/00029.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48781cf74b5aa8ff94627e10bba036e0348574f4ae7dfbe93c835e5aaf866908
+size 16899

audio/000/00030.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a3c6951c1ed4deb236c78c0b031e4f501cf1dede176e67a652a2ad0598aa472
+size 19564

audio/000/00031.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:928a42f68b4bec2d291f08fc4e42d4a3635bc8c8c66a8f4eb9dfe78c79852a6d
+size 12419

audio/000/00032.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:67cbc120429193f523fdbf544919fd067fec04fcf761fe6ea715c04c7c4c8b12
+size 18225

audio/000/00033.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2087a6587cb8a52ad016c212f1e758e898401d1cab2a1750e83666135ba643c3
+size 11162

audio/000/00034.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd457cf8ac193892d9196d9d1fb4b3fa0e875cb490a98836c36434094af24460
+size 9122

audio/000/00035.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bbc62d2341d0435f595e0c0ce897b4dffef9720d1ecc679cb23d70d41fa361c1
+size 17859

audio/000/00036.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:29fce3e283495c368abece7d14d0a0b22e86d07dfbf508c7d3b91d57f31a6ceb
+size 8530

audio/000/00037.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6666e4803ca8a0a21d8c6af33a12c4a322e1e03aedc70e137f9739f4c606622
+size 9228

audio/000/00038.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61b529a889184b32b6c46d87e5a955504de0fa6a4514a7e101f983ea43c5baa5
+size 16220

audio/000/00039.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5dce9832e95e9f96dcc446af6bb01053f0f87b690fd11daf370a1e3f6c25ed6
+size 15732

audio/000/00040.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5374c8695e6dda84831de8df1c1e8ac2469a81d284950d0e61934ba8934fc97b
+size 26366

audio/000/00041.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3822ca79065323c91a17f63f7c1c3a31b8645d22c81591e2b3775ae0f9f49d67
+size 13763

audio/000/00042.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:38a62d9b99e20cada01ef62782196e1369f2116344db649ec9d15d010f3ef517
+size 20346

audio/000/00043.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:448f6d6933775cd5bd20e4b9a125167eead79e03f27e320e4c549a72f8c998f6
+size 16944

audio/000/00044.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ee95c8d762721b198a906dcdd16a4151bb9adcc856ac958dfcd07cad0d7f3bc
+size 13358

audio/000/00045.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:baa3039e02eac3aa2aee0b39403f2b4a91d3453363980cee079c386978c4bab2
+size 12249

audio/000/00046.ogg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3cd3d5303fa720944cb22ec1656b125678c603edbde7be53bd5dbfc7e916be16
+size 10556