Uninorm operators for sentence-level score aggregation in sentiment analysis

With the rapid growth of the Web and its users around the world, the availability and usefulness of users' comments have increased in recent years. Sentiment analysis techniques are used to extract valuable knowledge from this increasing volume of textual information. These techniques usually decompose every comment to its sentences, detect their sentiments, and aggregate these sentiments to calculate the overall sentiment being conveyed in the comment. In this study, a new sentence-level aggregation mechanism based on uninorm operators is proposed for aggregating sentence-level sentiment into an overall document-level opinion. In order to show the utility of the proposed method, it is applied to polarity detection and score prediction problems on four large Persian review datasets. Implementation results show that the proposed method, in comparison with Dempster-Shafer aggregation method, achieves a higher performance in polarity detection while the Dempster-Shafer method slightly outperforms the proposed method in score prediction task.