Detecting malicious tweets in trending topics using clustering and classification

Soman, Saini Jacob; Murugappan, S.

doi:10.1109/icrtit.2014.6996188

Cited by 17 publications

(1 citation statement)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In another study done in paper [15] the problem of feature extraction was tried to be solved. For performing this, in this work, initially different features such as user activity features, location based features, content and text features and user pro le features were extracted.…”

Section: Related Workmentioning

confidence: 99%

Topic extraction and categorization using a glove based technique for representing topics

Scholar,

Hazarika,

Mahanta

2023

Preprint

View full text Add to dashboard Cite

Topic extraction and categorization is an important task because by doing that it is easy to find out which are the topics most discussed by the users in their tweets or opinions and need to be analyzed. In this work, topics are extracted from positive and negative opinions and then categorized into different groups. For performing this, first a collection of opinions are divided into two sets- positive opinions and negative opinions by using a sentiment analyzer. Then a method is proposed to find out the most discussed topics in the set of positive opinions and negative opinions. For extracting the topics from a set of opinions the noun words are extracted from the set of the opinions. After extracting the topics, the similar topics have been combined by using synonymy relation. Then the frequent topic words are represented with the help of GloVe embedding technique. Finally, the topics are categorized by using a clustering algorithm by applying it on the frequent topic words. For the evaluation of the proposed method, tweets from a Twitter User dataset are used. The results obtained from the experiments by applying the proposed method on the dataset give promising result and provide interesting and meaningful clusters of topics. Moreover an analysis of the result obtained for both positive and negative opinions is also presented.

show abstract

Section: Related Workmentioning

confidence: 99%