Deep learning for hate-speech detection in Afaan Oromo audio and video

In today world, identification of hate speech is difficult in audio and video content especially for understudied language like Afaan-Oromo. Since it is complicated language that requires efficient processing and analyzing. Developing deep learning approach for identifying hate speech in Afaan-Oromo language using audio and video content is presented in this work. we used different libraries such as librosa and movplay and spectrograms as input data for identifying hate speech in Afaan Oromo. Each library has their own function for example librosa library used to extract meaningful audio features from audio and video data, and transform them into spectrograms. After the model trained well we used a validation dataset to assess the of the model. The modelvalidation loss for starting was 59.34, indicating some degree of prediction error. Also, it achieved a validation accuracy of 68.72, indicating its effectiveness in accurately categorizing hate speech. However, by investigating cutting-edge deep learning methods and expanding the dataset with a variety of hate speech, more advancement are required to increase the modelperformance.The suggested model demonstrated encouraging outcomes after being trained on a CNN utilising spectrograms. In this study data which only contained voice and video were used. We required Afaan Oromo datasets to develop the model however they weren't available before and we prepared hate speech dataset in both audio and vidio. The main focus was on the content of the videos rather than image they showed.