Automating the extraction of otology symptoms from clinic letters: a methodological study using natural language processing

Nikhil Joshi ^1,2, Kawsar Noor ^3,4, Xi Bai ^3,4

⁰evidENT Team, Ear Institute, UCL, London, UK. n.joshi@ucl.ac.uk.

Bmc Medical Informatics and Decision Making +

|

September 30, 2025

View abstract on PubMed

Summary

This study developed natural language processing (NLP) models to automatically extract and contextualize otology symptoms from clinical notes. These models show promise for improving healthcare research and electronic health record analysis.

Area Of Science

Medical Informatics
Computational Linguistics
Otolaryngology Research

Background

Healthcare data is predominantly unstructured, hindering research scalability.
Manual data processing is time-consuming and inefficient.
Natural Language Processing (NLP) offers automated solutions for data extraction.

Purpose Of The Study

To develop NLP models for extracting and contextualizing otology symptoms from free-text clinical documents.
To automate the processing of unstructured otology-related healthcare data.

Main Methods

A hybrid dictionary and machine learning NLP model was trained on 1,148 otology clinic letters.
Bidirectional-Long-Short-Term-Memory (Bi-LSTM) models were used for symptom contextualization.
Six key otological symptoms were targeted: hearing loss, balance impairment, otalgia, otorrhoea, tinnitus, and vertigo.

Main Results

The symptom extraction model achieved a macro F1 score of 0.73.
Bi-LSTM models achieved a mean macro F1 score of 0.69 for contextualization.
24% of patients presented with hearing loss, with 1,197 symptom and 2,861 contextual annotations.

Conclusions

Successfully developed NLP models for otology symptom extraction and contextualization.
Models demonstrate good performance on real-world clinical data.
Potential applications include semantic searching, clinical trial cohort identification, and hearing loss phenotype research.

Keywords:

Machine learning Natural language processing Otology Symptoms