Belay, Ayalew (PhD)Wonago, Hiwot2020-08-192023-11-042020-08-192023-11-047/17/2020http://etd.aau.edu.et/handle/123456789/22093In the last five years, the ever growing usage of social media in Ethiopia has fueled the country‘s problem against the peaceful coexistence of its people. Illegitimate social media usage has played a significant role in widening the distress between the people. As a result, the government has increasingly relied on the temporary closure of social media sites; nationwide internet shutdowns and filtering websites to suppress polarizing voices and the misuse of social media as the tension among many ethnic groups become more visible. As such, there is a need to develop an intelligent system that automatically detects such inappropriate (offensive) contents by classifying them into socially-offensive, religiouslyoffensive, politically-offensive and non-offensive categories and filter Toxic online contents. We explain the challenges of the Amharic text that is available on the internet and the role of sentiment analysis in mining Amharic dataset on social media. Using different supervised machine learning techniques, this study analyzed performance variations of the algorithms on Amharic texts. The objective of this paper is to apply the concept of sentiment analysis on Amharic text on social media and presents a comparative study on machine learning algorithms. The created social media content filtering system has been tested on Facebook posts of each class, and it has been observed that SVM with word2vec has performed best in comparison to other classifiers, achieving average precision of (72%), but did worse on recall(63.4%). The experimental evaluation shows how the proposed approach is effective and the results are quite satisfactory.enInformation FilteringNatural Language ProcessingSentiment ClassificationWord2vecThesis