Offensive Language Detection in Nepali Social Media

Abstract

Social media texts such as blog posts, comments, and tweets often contain offensive languages including racial hate speech comments, personal attacks, and sexual harassments. Detecting inappropriate use of language is, therefore, of utmost importance for the safety of the users as well as for suppressing hateful conduct and aggression. Existing approaches to this problem are mostly available for resource-rich languages such as English and German. In this paper, we characterize the offensive language in Nepali, a low-resource language, highlighting the challenges that need to be addressed for processing Nepali social media text. We also present experiments for detecting offensive language using supervised machine learning. Besides contributing the first baseline approaches of detecting offensive language in Nepali, we also release human annotated data sets to encourage future research on this crucial topic.

Multi-Aspect Annotation and Analysis of Nepali Tweets on Anti-Establishment Election Discourse

“…They employed various machine learning and deep learning methods to perform benchmark classifications. Contributing more with respect to sentiment analysis, Niraula et al [12] collected 7,462 records (comments, posts, articles) from various social media platforms like Facebook, Twitter, YouTube, Blogs, and News Portals, and annotated them for sentimental analysis. Their analysis primarily focused on offensive language detection.…”

“…Over time, the field of NLP in Nepali has witnessed certain advancements. However, given the morphologically rich nature of the language [12] and the complex sentence structure [16], NLP in Nepali is particularly challenging and demands vigorous research. NLP advancements in any low-resource language like Nepali are usually restricted by the lack of pretraining data, resource uniformity, and computing resources [35].…”

“…Nepalithe most spoken language in Nepal [11], is also the major language for social media discourse in Nepal. The complex and monographically rich nature of the Nepali language makes Natural Language Processing (NLP) tasks particularly challenging for this language [12]. Unlike languages such as English, which has ample resources and a plethora of NLP studies [13], [14], research in the field of low-resource languages such as Nepali is scarce [15], [16].…”

“…Unlike languages such as English, which has ample resources and a plethora of NLP studies [13], [14], research in the field of low-resource languages such as Nepali is scarce [15], [16]. There have been some works involving sentimental analysis in Nepali [12], [17], [18]. These works have focused on a single aspect of sentiment analysis such as hate speech detection or abuse analysis.…”

NEHATE: Large-Scale Annotated Data Shedding Light on Hate Speech in Nepali Local Election Discourse

“…The M-BERT model's performance was found to be inadequate due to the limited size of Wikipedia content available for low-resource languages such as Nepali, which was used for training. NLP research in a morphologically rich and complex language like Nepali [17], poses several challenges. One of the major challenges is the sentence structure of the Nepali language [31].…”