Advancements in Speech Recognition and Medical Dialogue Systems

The field of speech recognition and medical dialogue systems is moving towards more accurate and informative models. Recent developments have focused on improving the distinction between repetition disfluency and morphological reduplication in low-resource languages, as well as advancing conversational speech recognition in underrepresented languages. Additionally, there is a growing interest in using large language models to simulate standardized patients for medical education and to assess the clinical impact of transcription errors in patient-facing dialogue. Noteworthy papers include: Distinguishing Repetition Disfluency from Morphological Reduplication in Bangla ASR Transcripts, which introduces a novel corpus and benchmarking analysis for Bangla speech recognition. Toward Conversational Hungarian Speech Recognition, which introduces two new datasets for Hungarian speech recognition and establishes reproducible baselines. Human or LLM as Standardized Patients, which presents a multi-agent framework for simulating standardized patients and introduces a benchmark for evaluating their performance. WER is Unaware, which challenges the standard evaluation metric for ASR systems and introduces a new framework for assessing the clinical impact of transcription errors.

Sources

Distinguishing Repetition Disfluency from Morphological Reduplication in Bangla ASR Transcripts: A Novel Corpus and Benchmarking Analysis

Toward Conversational Hungarian Speech Recognition: Introducing the BEA-Large and BEA-Dialogue Datasets

Human or LLM as Standardized Patients? A Comparative Study for Medical Education

WER is Unaware: Assessing How ASR Errors Distort Clinical Understanding in Patient Facing Dialogue

Built with on top of