A Study on Multiword Expression Features in Emotion Detection of Code-Mixed Twitter Data

Abstract

The need for automated detection of emotions in microblogs such as Twitter is growing as many organizations realize the need to take the emotions of their customers and the public in general into consideration in their problem solving and decision making. In multilingual countries such as Malaysia, tweets are typically written in code-mixed Malay-English with some informal Romanized dialect texts. These tweets are often embedded with multiword expressions. Although there have been studies on multiword expressions, there has not been substantial work on the use of multiword expressions as features to detect emotion of text. This paper studies the use of multiword expressions extracted from WordNet and WordNet Bahasa as features in the emotion detection of code-mixed Twitter data.