تقرير
A Large-Scale Evaluation of Speech Foundation Models
| العنوان: | A Large-Scale Evaluation of Speech Foundation Models |
|---|---|
| المؤلفون: | Yang, Shu-wen, Chang, Heng-Jui, Huang, Zili, Liu, Andy T., Lai, Cheng-I, Wu, Haibin, Shi, Jiatong, Chang, Xuankai, Tsai, Hsiang-Sheng, Huang, Wen-Chin, Feng, Tzu-hsun, Chi, Po-Han, Lin, Yist Y., Chuang, Yung-Sung, Huang, Tzu-Hsien, Tseng, Wei-Cheng, Lakhotia, Kushal, Li, Shang-Wen, Mohamed, Abdelrahman, Watanabe, Shinji, Lee, Hung-yi |
| سنة النشر: | 2024 |
| المجموعة: | Computer Science |
| مصطلحات موضوعية: | Electrical Engineering and Systems Science - Audio and Speech Processing, Computer Science - Computation and Language, Electrical Engineering and Systems Science - Signal Processing |
| الوصف: | The foundation model paradigm leverages a shared foundation model to achieve state-of-the-art (SOTA) performance for various tasks, requiring minimal downstream-specific modeling and data annotation. This approach has proven crucial in the field of Natural Language Processing (NLP). However, the speech processing community lacks a similar setup to explore the paradigm systematically. In this work, we establish the Speech processing Universal PERformance Benchmark (SUPERB) to study the effectiveness of the paradigm for speech. We propose a unified multi-tasking framework to address speech processing tasks in SUPERB using a frozen foundation model followed by task-specialized, lightweight prediction heads. Combining our results with community submissions, we verify that the foundation model paradigm is promising for speech, and our multi-tasking framework is simple yet effective, as the best-performing foundation model shows competitive generalizability across most SUPERB tasks. For reproducibility and extensibility, we have developed a long-term maintained platform that enables deterministic benchmarking, allows for result sharing via an online leaderboard, and promotes collaboration through a community-driven benchmark database to support new development cycles. Finally, we conduct a series of analyses to offer an in-depth understanding of SUPERB and speech foundation models, including information flows across tasks inside the models, the correctness of the weighted-sum benchmarking protocol and the statistical significance and robustness of the benchmark. Comment: The extended journal version for SUPERB and SUPERB-SG. Published in IEEE/ACM TASLP. The Arxiv version is preferred |
| نوع الوثيقة: | Working Paper |
| URL الوصول: | http://arxiv.org/abs/2404.09385 |
| رقم الأكسشن: | edsarx.2404.09385 |
| قاعدة البيانات: | arXiv |
| الوصف غير متاح. |