SUMMARY
In Ethiopia, the problem of hate speech posts on social media has become challenging recently. Detecting hate speech posts on social media is a tedious and complex task due to the unstructured format of social media content, which requires some detection mechanisms. Due to the success of deep learning algorithms in natural language processing tasks, some researchers used deep learning models for hate speech detection. But, most of the existing studies are explored only for high resource languages like English, except some studies recently proposed also for low resource languages. This study proposedbilingual hate speech detection for Afaan Oromo and Amharic texts on social media using deep learning. A bilingual dataset prepared from newly collected Afaan Oromo texts from Facebook platform and the existing binary Amharic dataset is adopted to develop models. The prepared dataset contains binary classes “Hate” and “Free”. Bidirectional RNNs and attention mechanisms are implemented using Word2vec as feature representation. The word2vec model is trained based on the skip-gram model. The models are trained using 5-fold and also 10-fold cross-validation. The results show that, models achieved a good performance when using 5-fold cross-validation on our dataset. Then, several experiments are employed to select the best-performing model, and finally, the BiLSTM model outperformed all other models with an accuracy of 94.3% and f1_score of 94.2%.