Trend Detection in Sinhala Tweets Using Clustering and Ranking Algorithms

摘要

The huge popularization of social media has resulted in a large amount of information including news, gossip, events, entertainment, and political and religion-based information being shared among the users. Among such shared information, some information becomes more popular as it is frequently shared on social media platforms. A trending topic means an item that contains hot news or a subject that experiences huge popularity on one or more social media platforms within a specific limited duration of time. Even though there are numerous ways to detect trending topics in English, there is a dearth of studies on Sinhala trend detection. In this research, an approach is proposed to detect trending topics in Sinhala. Among many social media platforms, Twitter is a very popular microblogging platform. As such, it was selected for this study. Identification of trending topics on Twitter is very useful for parties who are doing business and related activities. The identification of trending topics is done through the clustering and ranking of tweets. Agglomerative hierarchical clustering was used for clustering, and a customized pairwise comparison algorithm was used for ranking. An overall accuracy of 69.29% was obtained via a qualitative evaluation. Textual features as well as non-textual features extracted from Tweets were used for the analysis.