The Generation of a Corpus for Clinical Sentiment Analysis

Yihan Deng, Thierry Declerck, Piroska Lendvai, Kerstin Denecke

In: Harald Sack , Giuseppe Rizzo , Nadine Steinmetz , Dunja Mladenić , Sören Auer , Christoph Lange (Hrsg.). The Semantic Web -- ESWC 2016 Satellite Events. Seiten 311-324 Lecture Notes in Computer Science (LNCS) 9989 ISBN 978-3-319-47601-8 Springer Cham 10/2016.


Clinical care providers express their judgments and observations towards the patient status in clinical narratives. In contrast to sentiment expressions in general domains targeted by language technology, clinical sentiments are influenced by related medical events such as clinical precondition or outcome of a treatment. We argue that patient status in terms of positive, negative and neutral judgements can only suboptimally be judged with generic approaches, and requires specific resources in term of a lexicon and training corpus targeting clinical sentiment. To address this challenge, we manually developed a corpus based on 300 ICU nurse letters derived from a clinical database, and an annotation scheme for clinical sentiment. The paper discusses influence patterns between clinical context and clinical sentiments as well as a semi-automatic method to generate a larger annotated corpus.

Weitere Links

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence