Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Extracting Biomedical Entities from Noisy Audio Transcripts

Nima Ebadi, Kellen Morgan, Adrian Tan, Billy Linares, Sheri Osborn, Emma Majors, Jeremy Davis, Anthony Rios

TL;DR

A novel dataset is introduced, BioASR-NER, designed to bridge the ASR-NLP gap in the biomedical domain, focusing on extracting adverse drug reactions and mentions of entities from the Brief Test of Adult Cognition by Telephone (BTACT) exam.

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 have primarily studied ASR's efficiency in clean recordings, leaving a research gap concerning the performance in noisy environments. This paper introduces a novel dataset, BioASR-NER, designed to bridge the ASR-NLP gap in the biomedical domain, focusing on extracting adverse drug reactions and mentions of entities from the Brief Test of Adult Cognition by Telephone (BTACT) exam. Our dataset offers a comprehensive collection of almost 2,000 clean and noisy recordings. In addressing the noise challenge, we present an innovative transcript-cleaning method using GPT4, investigating both zero-shot and few-shot methodologies. Our study further delves into an error analysis, shedding light on the types of errors in transcription software, corrections by GPT4, and the challenges GPT4 faces. This paper aims to foster improved understanding and potential solutions for the ASR-NLP gap, ultimately supporting enhanced healthcare documentation practices.

Extracting Biomedical Entities from Noisy Audio Transcripts

TL;DR

A novel dataset is introduced, BioASR-NER, designed to bridge the ASR-NLP gap in the biomedical domain, focusing on extracting adverse drug reactions and mentions of entities from the Brief Test of Adult Cognition by Telephone (BTACT) exam.

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 have primarily studied ASR's efficiency in clean recordings, leaving a research gap concerning the performance in noisy environments. This paper introduces a novel dataset, BioASR-NER, designed to bridge the ASR-NLP gap in the biomedical domain, focusing on extracting adverse drug reactions and mentions of entities from the Brief Test of Adult Cognition by Telephone (BTACT) exam. Our dataset offers a comprehensive collection of almost 2,000 clean and noisy recordings. In addressing the noise challenge, we present an innovative transcript-cleaning method using GPT4, investigating both zero-shot and few-shot methodologies. Our study further delves into an error analysis, shedding light on the types of errors in transcription software, corrections by GPT4, and the challenges GPT4 faces. This paper aims to foster improved understanding and potential solutions for the ASR-NLP gap, ultimately supporting enhanced healthcare documentation practices.
Paper Structure (20 sections, 4 figures, 4 tables)

This paper contains 20 sections, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Biomedical NER
  4. Biomedical ASR-NLP
  5. Data
  6. Dataset Curation
  7. CADEC.
  8. Synthetic BTACT.
  9. Audio Recordings and mixing
  10. Methods
  11. NER baselines.
  12. ASR.
  13. Zero-shot Prompting.
  14. Few-shot In-Context Learning.
  15. Results
  16. ...and 5 more sections

Figures (4)

  • Figure 1: Overview of our data collection process. The process has four main steps: 1) We collect the initial datasets (CADEC and BTACT); 2) Graduate assistants read and record the text in the datasets; 3) We normalize each audio file to the same loudness; and 4) we generate noisy audio files by merging multiple speakers and adding background noise.
  • Figure 2: Overview of the training procedure and the prediction strategies we explore to improve biomedical NER performance.
  • Figure 3: Example of the zero-shot prompting strategy we use for GPT4.
  • Figure 4: Example of the few-shot prompting strategy we use for GPT4.