IDENTIFYING HATE SPEECH IN AFAAN OROMO AUDIO AND VIDEO USING DEEP LEARNING APPROACH

The identification of hate speech in Afaan Oromo audio and video content is a challenging task due to the complex nature of the language and the need for efficient processing of audio and video data. In this study, we propose a deep learning approach for identifying hate speech in Afaan Oromo using the librosa and movplay libraries, along with spectrograms as input features. To train our model, we utilized a Convolutional Neural Network (CNN) architecture. The librosa library enabled us to extract meaningful audio features from the Afaan Oromo audio and video data, and transform them into spectrograms. Spectrograms provide a visual representation of the frequency and amplitude characteristics of the audio signal over time, allowing the model to capture discriminatory patterns and features associated with hate speech. The movplay library played a crucial role in handling both audio and video data during the training process. It facilitated seamless integration of the spectrograms into the CNN model pipeline, enabling efficient data augmentation, preprocessing, and model training. After training the model, we evaluated its performance using a validation dataset. The model achieved a validation loss of 61.4, indicating a reasonable level of error in its predictions. Moreover, it achieved a validation accuracy of 64.1, demonstrating its effectiveness in accurately classifying hate speech in Afaan Oromo audio and video content. However, further improvements are desired to enhance the model's performance. This may involve fine-tuning the CNN model architecture, exploring advanced deep learning techniques, and expanding the dataset with diverse and representative examples of hate speech in Afaan Oromo. Our study presents a deep learning approach for identifying hate speech in Afaan Oromo audio and video content. The proposed model, trained on a CNN architecture using spectrograms generated by librosa and movplay libraries, achieved promising results with a validation loss of 61.4 and validation accuracy of 64.1. This research contributes to addressing the challenge of hate speech detection in Afaan Oromo and lays the foundation for further advancements in this field.