A Symbolically Relevance Words Approach in Hash tag Graph-Based Model for Micro blog Sites

Abstract

In this paper, we have proposed to introduce a brand new topic model to know the chaotic microblogging atmosphere by exploitation hashtag graphs. Inferring topics on Twitter becomes has become significant however difficult task in several important applications may be. The shortness and informality of tweets ends up in extreme thin vector representations with an outsized vocabulary. This makes the standard topic models (e.g., Latent Dirichlet Allocation [1] and Latent linguistics Analysis [2]) fail to be told prime quality topic structures. Tweets area unit perpetually revelation with made user-generated hashtags. The hashtags create tweets semi-structured within and semantically associated with one another. Since hashtags area unit used as keywords in tweets to mark messages or to make conversations, they supply an extra path to attach semantically connected words. during this paper, treating tweets as semi-structured texts, we have a tendency to propose a unique topic model, denoted as Hashtag Graph-based Topic Model (HGTM) to find topics of tweets. By utilizing hashtag relation data in hashtag graphs, HGTM is in a position to find word linguistics relations even though words aren't co-occurred at intervals a particular tweet. With this methodology, HGTM with success alleviates the spareness drawback. Our investigation illustrates that the user-contributed hashtags might function weakly-supervised data for topic modeling, and also the relation between hashtags might reveal latent linguistic relation between words. we have a tendency to value the effectiveness of HGTM on tweet (hashtag) bunch and hashtag classification issues. Experiments on 2 real-world tweet information sets show that HGTM has sturdy capability to handle scantiness and noise drawback in tweets. Moreover, HGTM will discover a lot of distinct and coherent topics than the progressive baselines.