Publications/Speech Processing, Recognition & Audio Security
Conference · 2018
Rapid Collection of Spontaneous Speech Corpora using Telephonic Community Forums
Agha Ali Raza, Awais Athar, Shan Randhawa, Zain Tariq, Muhammad Bilal Saleem, Haris Bin Zia, Umar Saif, Roni Rosenfeld
Interspeech 2018, Hyderabad, India · CORE A
Summary
Building speech technology for a language takes many hours of natural speech, which is costly to collect. We show that a voice-based community forum, where people post, comment, vote, and share audio over ordinary phone calls, gathers it as a by-product. In 8 months the forum reached 11,017 users in Pakistan and collected 1,207 hours of spontaneous speech in Urdu and 9 regional languages. A recognizer trained on just 9.5 hours of it achieved a 24.19% word error rate. The approach takes care of speaker recruitment, informed consent, and content diversity, captures real-world background noise, and can help preserve endangered languages and oral cultures.
Cite this paper
Agha Ali Raza, Awais Athar, Shan Randhawa, Zain Tariq, Muhammad Bilal Saleem, Haris Bin Zia, Umar Saif, Roni Rosenfeld. Rapid Collection of Spontaneous Speech Corpora using Telephonic Community Forums. Interspeech 2018, Hyderabad, India.
@inproceedings{raza2018rapid,
author = {Agha Ali Raza and Awais Athar and Shan Randhawa and Zain Tariq and Muhammad Bilal Saleem and Haris Bin Zia and Umar Saif and Roni Rosenfeld},
title = {{Rapid Collection of Spontaneous Speech Corpora using Telephonic Community Forums}},
booktitle = {Interspeech 2018, Hyderabad, India},
year = {2018},
url = {https://aghaaliraza.com/publications/2018--INTERSPEECH--Rapid_Collection_of_Spontaneous_Speech_Corpora_using_Telephonic_Community_Forums.pdf}
}
- Dissecting ASR Failures in Low-Resource South Asian LanguagesInterspeech 2026, Sydney, Australia
- WER We Stand: Benchmarking Urdu ASR ModelsCOLING 2025, Abu Dhabi, UAE
- Conversations in the Wild: Data Collection, Automatic Generation and EvaluationComputer Speech & Language, vol. 89, 2025
- Deepfake Defense: Constructing and Evaluating a Specialized Urdu Deepfake Audio DatasetFindings of ACL 2024, Bangkok, Thailand
- Speech Emotion Recognition for the Urdu LanguageLanguage Resources and Evaluation (LRE), 2023