Submit Manuscript  

Article Details


Understanding Twitter Hashtags from Latent Themes Using Biterm Topic Model

Author(s):

Muzafar Rasool Bhat*, Burhan Bashir , Majid A. Kundroo and Naffi A. Ahanger   Pages 1 - 9 ( 9 )

Abstract:


Social media in general and Twitter in particular provides a space for discourses, contemporary narratives besides a discussion about few specific social issues. People respond to these events by writing short text messages.

Background: Hashtag “#”, a specific way to respond to a given raised discourse, narrative or any contemporary issue is usual to social media. Netizens write a short message as their opinion about any given issue represented using a given Hashtag. These small messages generally tend to have a latent topic (theme) as one’s opinion about it.

Objective: This research is aimed to extract, represent and understand those hidden themes

Method: Biterm Topic Model (BTM) has been used in this study given its ability to deal with the short messages unlike Latent Dirichlet Allocation that expects a document to have a significant length.

Results: Twitter Hashtag #MeToo has been used in this research with forty thousand (40,000) comments. Data has been modelled with ten (10) topics after verifying suitable number of topics from four metrics Griffths, CaoJuan, Arun and Deveaud.

Conclusion: The experimental results show that the proposed approach to understand the twittter hashtages from latent themes using biterm topic modelling method is very effective as compared to other methods

Keywords:

Topic Modelling, LDA, BTM, Social Media Analysis (SMA), twitter analysis, #MeToo.

Affiliation:

Department of Computer SciencesIslamic University of Science and Technology Awantipora J&K-192122;, Department of Computer SciencesIslamic University of Science and Technology Awantipora J&K-192122;, Department of Computer SciencesIslamic University of Science and Technology Awantipora J&K-192122;, Department of Computer SciencesIslamic University of Science and Technology Awantipora J&K-192122;



Read Full-Text article