Improving sentiment analysis on clinical narratives by exploiting UMLS semantic types

作者：

Highlights：

• Meanings of medical terms are useful for sentiment analysis on clinical narratives.

• Use of UMLS semantic types can improve lexicon-based sentiment analysis.

• Classifier combination tends to use a smaller training set than a single classifier.

• Not all UMLS semantic types in the ’Disorders’ group indicate negative sentiments.

摘要

•Meanings of medical terms are useful for sentiment analysis on clinical narratives.•Use of UMLS semantic types can improve lexicon-based sentiment analysis.•Classifier combination tends to use a smaller training set than a single classifier.•Not all UMLS semantic types in the ’Disorders’ group indicate negative sentiments.

论文关键词：Lexicon-based sentiment analysis,Clinical narrative,Unified medical language system,Classifier combination,Domain-specific knowledge

论文评审过程：Received 3 March 2020, Revised 26 January 2021, Accepted 9 February 2021, Available online 12 February 2021, Version of Record 26 February 2021.

论文官网地址：https://doi.org/10.1016/j.artmed.2021.102033