HATE SPEECH IDENTIFICATION IN SOCIAL MEDIA USING  SENTIMENT ANALYSIS FOR AMHARIC LANGUAGE

TEGEGN, SEYFE ASSEFA

dc.contributor.author	TEGEGN, SEYFE ASSEFA
dc.date.accessioned	2023-06-19T12:06:29Z
dc.date.available	2023-06-19T12:06:29Z
dc.date.issued	2022-11
dc.identifier.uri	http://ir.bdu.edu.et/handle/123456789/15406
dc.description.abstract	Hate speech is a critical problem in Ethiopia and around the world and is disseminated using different social media platforms. In the world, as social media users increase, hate speech dissemination increases rapidly in parallel. To overcome the dissemination of hate speech, social media platforms have the responsibility of controlling users. In Ethiopia, to control hate speech dissemination, a hate speech proclamation has been declared. However, the dissemination of hate speech is still a critical problem in the country. In this research, to help limit hate speech dissemination, we build an automatic hate speech detection model. The main focus of the research is building a hate speech detection model using sentiment analysis and demonstrating a correlation between sentiment analysis and hate speech. We develop a hate speech detection model for the Amharic language using machine learning and deep learning techniques. Moreover, we investigate the impact of sentiment analysis on hate speech on social media. We find out that most of the hate speech content originated from negative sentiment. The experiment is performed using both machine learning and deep learning approaches. In the experiment with human-annotated datasets, Random Forest from the machine learning model and BiLSTM from the deep learning model performed with 0.79 and 0.82 accuracies respectively. While in the hate speech detection model trained, tested, and validated using both human-annotated and sentiment model-annotated datasets, LSTM and Bi-LSTM perform with 0.53 and 0.82 accuracies respectively. Hence, the hate speech detection model built by incorporating sentiment analysis has good performance using the BiLSTM approach. The main contributions of this work include the findings that the developing automatic hate speech detection model incorporates sentiment analysis, a multi-class labeled dataset, and a hate speech annotation guideline. Key Words: - Hate speech detection model, sentiment analysis, deep learning, Amharic language, hate speech, multi-tasking.	en_US
dc.language.iso	en_US	en_US
dc.subject	Computing	en_US
dc.title	HATE SPEECH IDENTIFICATION IN SOCIAL MEDIA USING SENTIMENT ANALYSIS FOR AMHARIC LANGUAGE	en_US
dc.type	Thesis	en_US