Improving the performance of sentiment analysis in health and wellbeing using domain knowledge

Zunic, Anastazia, Corcoran, Padraig ORCID: https://orcid.org/0000-0001-9731-3385 and Spasic, Irena ORCID: https://orcid.org/0000-0002-8132-3885 2020. Improving the performance of sentiment analysis in health and wellbeing using domain knowledge. Presented at: the Third UK Healthcare Text Analytics Conference (HealTAC), London, UK, 22-24 April 2020. -.

Full text not available from this repository.

Abstract

Sentiment analysis is a natural language processing task that aims to automatically classify the sentiment expressed in text. In this study, we compare the performance of five publicly available sentiment analysis tools along with the ensemble method that combines them. Their performance was evaluated on two datasets, which represent user-generated content. One of these, namely drug reviews, is related to health and wellbeing. The second one, movie reviews, is used for cross-domain comparison of sentiment analysis. Explicit domain knowledge formally modelled by the Unified Medical Language System was used for semantic enrichment to investigate whether it can improve the performance of the sentiment analysis tools considered, by reducing the bias towards the negative sentiment. Our experiments demonstrated an improvement in F-score by 7 percent points.

Item Type:	Conference or Workshop Item (Paper)
Status:	In Press
Schools:	Computer Science & Informatics Data Innovation Research Institute (DIURI)
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science Q Science > QA Mathematics > QA76 Computer software
Date of Acceptance:	26 February 2020
Last Modified:	06 Jan 2024 02:29
URI:	https://orca.cardiff.ac.uk/id/eprint/129997

Actions (repository staff only)

Edit Item