A two-stage contagious Naive Bayes classifier for detecting sociolinguistic features in text

Loading...
Thumbnail Image

Date

Authors

Derks, Iena Petronella
De Waal, Alta

Journal Title

Journal ISSN

Volume Title

Publisher

CEUR Workshop Proceedings

Abstract

Online platforms allow users to masquerade themselves; making virtual interactions anonymous or misleading recipients of the interactions. It also facilitates an environment for cybercrimes, allowing users to take advantage of others and commit heinous acts. An important concern on social media usage, in particular, has to do with the security of under-age users that have access to the Internet. Children are more vulnerable to threatening situations, such as harassment, cyberbullying, and inappropriate conversations. Natural language processing (NLP) techniques can be used to process and understand social media data. In the area of sociolinguistics, there is evidence that links natural word use to personality and social fluctuations. In NLP, the term burstiness is used to describe the tendency of word recurrence. The burstiness phenomenon is frequently exhibited in real text, in which an informative word is more likely to occur if it has already appeared in the text. State-of-the-art NLP models, such as the multinomial Naive Bayes model, are often used to model text documents.

Description

Keywords

Online platforms, Cybercrimes, Naive Bayes model, Natural language processing (NLP)

Sustainable Development Goals

Citation

Derks, I.P. & De Waal, A. 2019, 'A two-stage contagious Naive Bayes classifier for detecting sociolinguistic features in text', CEUR Workshop Proceedings, vol. 2540, pp. 1-2.