Un-Compromised Credibility: Social Media based Multi-Class Hate Speech Classification for Text: A Review

Abstract

Hate speech is a crime that has been on the rise in recent years, not just in face-to-face contacts but also online. This is due to a number of causes. On the one hand, due of the anonymity given by the internet and social networks in particular, people are more likely to engage in hostile behaviour. People\'s desire to voice their thoughts online, on the other side, have increased, adding to the spread of hate speech. Governments and social media platforms can benefit from detection and prevention techniques because this type of prejudiced speech can be immensely destructive to society. We contribute to a solution to this dilemma by giving a systematic review of research undertaken in the subject through this survey. This challenge benefited from the use of several complicated and non-linear models, and CAT Boost performed best due to the application of latent semantic analysis (LSA) for dimensionality reduction.

Introduction

I. INTRODUCTION

Online social network (OSN) is the use of dedicated websites applications that allow users to interact with other users or to ?nd people with similar own interest Social networks sites allow people around the world to keep in touch with each other regardless of age [1] [7]. Sometimes children are introduced to a bad world of worst experiences and harassment. Users of social network sites may not be aware of numerous vulnerable attacks hosted by attackers on these sites. Today the Internet has become part of the people daily life. People use social networks to share images, music, videos, etc., social networks allows the user to connect to several other pages in the web, including some useful sites like education, marketing, online shopping, business, e-commerce and Social networks like Facebook, LinkedIn, Myspace, Twitter are more popular lately [8][9]. The offensive language detection is a processing activity of natural language that deals with ?nd out if there are shaming (e.g. related to religion, racism, defecation, etc.) present in a given document and classify the ?le document accordingly [1]. The document that will be classi?ed in abusive word detection is in English text format that can be extracted from tweets, comments on social networks, movie reviews, and political reviews. Hate speech is a crime that has been on the rise in recent years, not just in face-to-face contacts but also online. This is due to a number of causes. On the one hand, due of the anonymity given by the internet and social networks in particular, people are more prone to engage in hostile behaviour, People, on the other hand, are more willing to share their thoughts online, which contributes to the spread of hate speech as well. Governments and social media platforms can benefit from detection and prevention techniques because this type of prejudiced speech can be immensely destructive to society. We contribute to a solution to this dilemma by giving a thorough overview of research undertaken in this area through this survey.

Hate speech is defined as a discourse that is potentially hurtful to a person's or group's feelings and may contribute to violence or insensitivity, as well as irrational and inhuman behaviour. Hate speech has increased as a result of the growth of online social media, which is illegal. Hate speech and hate crimes are linked, and there is evidence that hate crimes are on the rise. As the problem of hate speech grows in popularity, many government-led initiatives are being implemented, such as the Council of Europe's No Hate Speech movement. Legislation has also been enacted to combat its spread, dubbed the EU Hate Speech Code of Conduct, which must be signed and implemented by all social media services within 24 hours.

The majority of the issues raised are primarily connected to the dataset's quality, which will be addressed in this study through the creation of quality-based strong datasets. The second problem, which is also addressed in this paper, is to investigate and determine the best set of characteristics for hate speech identification before developing a suitable classifier. When looking at the FBI's hate crime data, the most common categories are based on race, ethnicity, and religion. As a result, all of these categories are largely chosen for the production of datasets.

II. LITERATURE REVIEW

III. OPEN ISSUES

Lot of work has been done in this field because of its extensive usage and applications. In this section, some of the approaches which have been implemented to achieve the same purpose are mentioned. These works are majorly differentiated by the techniques for multi- keyword search and group sharing systems.

1. In previous technology in which A Survey on Automatic Detection of Hate Speech in Tex word sequence was ignored.
2. In White Supremacy Forum, The dataset is unbalanced as there exist many more sentences not conveying hate speech than ‘hateful” ones.
3. The Effects of User Features on Twitter Hate Speech Detection, this subset improvement may have been affected by the individual feature (number of) ‘Followers’, which also increased the F1-score on the two datasets.
4. The proposed sets of unigrams and patterns can be used as already-built dictionaries not included it is used for future works related to hate speech detection.

Conclusion

The complex problem of multi-class automated hate speech categorization for text is solved with considerably better results after the primary challenges are discovered first. There are ten unique binary categorised datasets made up of various hate speech categories. Experts annotated each dataset with a high level of agreement among annotators, using a set of detailed, well-defined guidelines. The datasets were well-balanced and comprehensive. They were also enriched with linguistic nuance. Compilation of such a dataset was accomplished as an essential need for filling the field\'s gap.

References

[1] Fortuna, Paula, and Sérgio Nunes. \"A survey on automatic detection of hate speech in text.\" ACM Computing Surveys (CSUR) 51.4 (2018): 1-30. [2] Kumar R, Ojha AK, Malmasi S, Zampieri M. Benchmarking Aggression Identification in Social Media. In: Proceedings of the First Workshop on Trolling, Aggression and Cyberbullying (TRAC-2018). ACL; 2018. p. 1–11. [3] de Gibert O, Perez N, Garc’ia-Pablos A, Cuadros M. Hate Speech Dataset from a White Supremacy Forum. In: 2nd Workshop on Abusive Language Online@EMNLP; 2018. [4] Davidson, Thomas, et al. \"Automated hate speech detection and the problem of offensive language.\" Proceedings of the International AAAI Conference on Web and Social Media. Vol. 11. No. 1. 2017. [5] Unsvåg, Elise Fehn, and Björn Gambäck. \"The effects of user features on twitter hate speech detection.\" Proceedings of the 2nd workshop on abusive language online (ALW2). 2018. [6] Vu, Xuan-Son, et al. \"HSD shared task in VLSP campaign 2019: Hate speech detection for social good.\" arXiv preprint arXiv:2007.06493 (2020). [7] Mathew, Binny, et al. \"Analyzing the hate and counter speech accounts on twitter.\" arXiv preprint arXiv:1812.02712 (2018). [8] Gaydhani, Aditya, et al. \"Detecting hate speech and offensive language on twitter using machine learning: An n-gram and tfidf based approach.\" arXiv preprint arXiv: 1809.08651 (2018). [9] Watanabe, Hajime, Mondher Bouazizi, and Tomoaki Ohtsuki. \"Hate speech on twitter: A pragmatic approach to collect hateful and offensive expressions and perform hate speech detection.\" IEEE access 6 (2018): 13825-13835. [10] Wich, Maximilian, Jan Bauer, and Georg Groh. \"Impact of politically biased data on hate speech classification.\" Proceedings of the Fourth Workshop on Online Abuse and Harms. 2020.

Copyright

Copyright © 2022 Miss. Priyanka R Telshinge, Mr. Mangesh D Salunke. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.