Summary of Xcb: An Effective Contextual Biasing Approach to Bias Cross-lingual Phrases in Speech Recognition, by Xucheng Wan et al.
XCB: an effective contextual biasing approach to bias cross-lingual phrases in speech recognition
by Xucheng Wan, Naijun Zheng, Kai Liu, Huan Zhou
First submitted to arxiv on: 20 Aug 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Sound (cs.SD); Audio and Speech Processing (eess.AS)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The study proposes a novel module, Cross-lingual Contextual Biasing (XCB), to improve automatic speech recognition (ASR) in bilingual settings. The XCB module integrates an auxiliary language biasing component and a supplementary language-specific loss into a pre-trained ASR model for the dominant language. This approach enhances the recognition of phrases in the secondary language without requiring additional inference overhead. Experimental results on an in-house code-switching dataset demonstrate significant improvements in recognizing biasing phrases, even surpassing benchmark performance. The proposed system also exhibits efficiency and generalization when applied to unseen datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper solves a problem with speech recognition by introducing a new way to make it better for people who speak multiple languages. Currently, speech recognition models are good at understanding one language but struggle when listening to speakers who switch between languages. The researchers created a special module that helps the model understand phrases in the second language, even if it hasn’t seen them before. They tested this on some data they made themselves and found that it works really well. This could be useful for applications like voice assistants or speech-to-text technology. |
Keywords
» Artificial intelligence » Generalization » Inference
