Linguistic Distance Segregates Latent Representations in Automatic Speech Recognition Systems

Language Discrimination Improves Linguistic Learning in Multilingual Speech Models

It is shown that strengthening the model's ability to discriminate languages during pretraining reduces and, on some measures, closes this multilingual gap on continuous phonetic and higher-level linguistic measures, while preserving substantial cross-language sharing.

Generative vs. Encoder Large Language Models for ASR Evaluation: A Comparative Study

The results show that encoder-based metrics remain highly competitive, while generative LLMs perform strongly in hypothesis comparison and improve the interpretability of ASR evaluation.

Do Speech Representations Preserve Regional Accent Across Read and Spontaneous Speech?

Regional accent cues can be captured under matched conditions, but it remains unclear whether they persist between read and spontaneous speech. We study RVG1, with 500 German speakers from nine regions, comparing ten speech representations on regional classification and continuous geolocation under matched conditions a...

Do speech foundation models really learn words?

It is shown that this simple approach to disentanglement can enhance higher-order linguistic information in word discovery tasks and learn representations which encode words with reasonable fidelity independently of local phonetic content.

Leveraging Linguistic Similarity for Low-Resource Speech Transcription

A Comprehensive Study of Content Representations for Speech Synthesis

Speech content representations are central to voice conversion, speech-to-speech translation, and multimodal language models, yet they are rarely compared under a common generative framework that directly measures what each representation contains. We address this by training a generative model conditioned solely on ea...