LID Models are Actually Accent Classifiers: Implications and Solutions for LID on Accented Speech
This addresses a critical issue for speech technology applications in multilingual and accented contexts, offering incremental improvements to existing LID systems.
The paper tackled the problem of language identification (LID) models performing poorly on accented speech by identifying that they often misclassify based on short phonotactic features indicative of accent rather than language, and proposed solutions like input chunking and integrating sequence-level information to reduce errors and enhance performance on accented speech while maintaining results on standard LID.
Prior research indicates that LID model performance significantly declines on accented speech; however, the specific causes, extent, and characterization of these errors remain under-explored. (i) We identify a common failure mode on accented speech whereby LID systems often misclassify L2 accented speech as the speaker's native language or a related language. (ii) We present evidence suggesting that state-of-the-art models are invariant to permutations of short spans of speech, implying they classify on the basis of short phonotactic features indicative of accent rather than language. Our analysis reveals a simple method to enhance model robustness to accents through input chunking. (iii) We present an approach that integrates sequence-level information into our model without relying on monolingual ASR systems; this reduces accent-language confusion and significantly enhances performance on accented speech while maintaining comparable results on standard LID.