arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2406.09282cs.CLcs.SDeess.AS
收藏

异构数据源对语音转文本基础模型的影响

On the Effects of Heterogeneous Data Sources on Speech-to-Text Foundation Models

Carnegie Mellon University(卡内基梅隆大学);Toyota Technological Institute at Chicago(芝加哥丰田理工学院)

Jinchuan Tian, Yifan Peng, William Chen, Kwanghee Choi, Karen Livescu, Shinji Watanabe

AI总结:

本文提出 OWSM v3.2,通过代理任务过滤数据并利用开放 LLM 恢复标点与真实大小写,在训练数据减少15%的情况下超越 OWSM v3.1。

摘要:

Open Whisper-style Speech Model(OWSM)系列的提出,是为了在构建先进语音转文本(S2T)基础模型时实现完全透明。为此,OWSM 模型在25个公开语音数据集上进行训练,这些数据集在多个方面具有异构性。在本研究中,我们通过推出 OWSM v3.2 推进了 OWSM 系列;该版本通过研究并解决这种数据异构性的影响,改进了以往模型。我们的研究首先对每个数据集进行详细分析,并由此得出两项关键策略:利用代理任务进行数据过滤以提升数据质量,以及使用开放的大型语言模型(LLM)加入标点符号和真实大小写。在所有其他配置保持不变的情况下,OWSM v3.2 使用的训练数据减少15%,同时性能优于 OWSM v3.1 基线。

英文摘要:

The Open Whisper-style Speech Model (OWSM) series was introduced to achieve full transparency in building advanced speech-to-text (S2T) foundation models. To this end, OWSM models are trained on 25 public speech datasets, which are heterogeneous in multiple ways. In this study, we advance the OWSM series by introducing OWSM v3.2, which improves on prior models by investigating and addressing the impacts of this data heterogeneity. Our study begins with a detailed analysis of each dataset, from which we derive two key strategies: data filtering with proxy task to enhance data quality, and the incorporation of punctuation and true-casing using an open large language model (LLM). With all other configurations staying the same, OWSM v3.2 improves performance over the OWSM v3.1 baseline while using 15% less training data.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑