Extracting Biomedical Entities from Noisy Audio Transcripts

arXiv:2403.17363v1 Announce Type: new
Abstract: Automatic Speech Recognition (ASR) technology is fundamental in transcribing spoken language into text, with considerable applications in the clinical realm, including streamlining medical transcription and integrating with Electronic Health Record (EHR) systems. Nevertheless, challenges persist, especially when transcriptions contain noise, leading to significant drops in performance when Natural Language Processing (NLP) models are applied. Named Entity Recognition (NER), an essential clinical task, is particularly affected by such noise, often termed the ASR-NLP gap. Prior works …

abstract applications arxiv asr audio automatic speech recognition biomedical challenges clinical cs.cl ehr electronic electronic health record health language medical natural noise performance recognition speech speech recognition spoken systems technology text transcription transcripts type