Leveraging Deep Learning in Hate Speech Analysis on Social Platform

Title

Leveraging Deep Learning in Hate Speech Analysis on Social Platform

Creator

Manikandan R.; Hariharasitaraman S.; Ramkumar S.; Gobinath R.

Description

The scope and usage of the Internet have surpassed the expected growth and have proven beyond the basic purpose of being used for networking and telecommunications. It serves as the backbone of the web, and one of the predominant domains that uses the Internet is social media. The concept was conceived in the early 1990s and went on to grow as a powerful medium of people networking along with the Internet. Social networking sites (SNS) acquired a predominant element of the Internet owing to their use and services they offer through the Internet. A few of the most used social networking sites include Twitter and Facebook, which are used synonymous to expressions of text. These SNS allow the users to post photos, videos, and other multimedia content along with text and voice messages that are shared among other users. As with any technology or application, these also have the risk of users posting offensive material and textual content. Hate is being spread through messages, which are in the form of text and also through other materials posted. There is no control to check for the message for the hate content as and when it is posted, and by the time it is deleted by admins, it could have already reached millions of users. This chapter proposes a technique for detecting hate texts in reviews from registered users in the Twitter dataset. The proposed work makes use of improved principle component analysis (IPCA) and modified convolution neural network (MCNN) for detecting hate texts. The advantage of natural language processing is used for building an automated system for the analysis of syntax and semantics of the words. The proposed methodology consists of phases like pre-processing, feature extraction, and process to classify the text. The white spaces in the text are removed through normalization in the pre-processing phase, and also remove special characters such as question marks, punctuations, and exclamatory symbols to remove stop words. The features that are pre-processed are then subjected to feature extraction using IPCA. A set of correlated features are made used for identifying more important features in the data set under consideration. Next, the classification is done for identifying the hate text or for any language abuse. MCNN is applied for the classification of the text into HATE and NON-HATE from the text with better accuracy. The experiments prove that the proposed method has a high level of accuracy even for a large dataset. The results show that the proposed method has better performance in terms of precision, recall, and F-measure when compared with other state-of-the-art methods. 2024 Taylor & Francis Group, LLC.

Source

Deep Learning for Smart Healthcare: Trends, Challenges and Applications, pp. 38-53.

Date

2024-01-01

Publisher

CRC Press

Coverage

Manikandan R., CHRIST (Deemed to be University), India, School of Business and Management, India; Hariharasitaraman S., Division of Cyber Security and Digital Forensics, Faculty of Computing Science and Engineering, VIT Bhopal University, India; Ramkumar S., Christ University, Bangalore, India; Gobinath R., Christ University, Bangalore, India

Rights

Restricted Access

Relation

ISBN: 978-104002137-8; 978-103245581-5

Format

Online

Language

English

Type

Book chapter

Identifier

https://doi.org/10.1201/9781003469605-3

https://www.scopus.com/inward/record.uri?eid=2-s2.0-85192317836&doi=10.1201%2f9781003469605-3&partnerID=40&md5=7d0447156d86c79b8dbb22c8cc9ebb1e