SENTIMENT CLASSIFICATION FOR SOCIAL MEDIA AMHARIC TEXT USING DEEP NEURAL NETWORK APPROACH: POLITICS DOMAIN

Social media is a tool that political parties utilize to genuinely communicate with the public about their objectives, strategies, and development policies. As a result, a large number of voters and social media users can take an active role by commenting or addressing such political concerns. Political parties can therefore use public feedback as a source of information when making decisions about their governmental policies and goals, taking into account the opinions and suggestions expressed by the general public.The manual analysis of large-scale political comments and posts is both challenging and timeconsuming. Existing sentiment classification studies for Amharic texts are limited, particularly in the political context, creating a need for advanced computational approaches.   The main goal of this work is deep learning approaches to classify Amharic social media political posts and comments into Positive, Negative, and Neutral, making the sentiment classification problem effective. dataset was then analyzed and modeled using deep learning techniques like Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Bidirectional Long Short-Term Memory (BiLSTM) to develop a sentiment classification model. Furthermore, in an effort to enhance classification performance, word embedding techniques like Word2Vec and FastText were investigated. We examined a number of deep learning algorithms and word embedding methods to see which was the best. By using BiLSTM with FastText embedding, the suggested approach was able to attain improved accuracy, precision, recall, and F1-score with 0.84, 0.92, 0.86, and 0.89, respectively. This provides a valuable tool for political parties and analysts, enabling efficient sentiment analysis of social media content, which can significantly influence policy formulation and political strategies.