Afaan Oromoo language Audio Sentiment Analysis Research Paper

This research delves into the linguistic and technological facets of the Afaan Oromo language,emphasizing its foundational elements, writing systems, and phonetic characteristics. The studyfurther extends to sentiment analysis through two main experiments: raw audio data analysis andtext-based analysis. These experiments are pivotal for understanding sentiment expression in AfaanOromo, combining linguistic insights with advanced deep learning techniques.The research methodology follows the Design Science Research Process (DSRP), addressingchallenges in sentiment analysis through a systematic approach. The study includes a comprehensivedataset comprising 1000 labeled audio samples and 7000 labeled text samples in Afaan Oromo. Twomain experiments are conducted, utilizing Convolutional Neural Networks (CNNs) for audioprocessing and exploring pretrained models for text-based sentiment analysis.In the analysis of audio data, CNNs prove effective in capturing local patterns, achieving an accuracyof 90.41% and an F1 score of 90.59. Audio preprocessing methods, including spectrogram generationand data augmentation, contribute to the model's robustness. The text-based analysis, however, yieldsinsights into the challenges of transcription models, with performance influenced by both the modeland transcription accuracy.The findings underscore the importance of considering data nature and context in sentiment analysis,suggesting potential avenues for future research, such as hybrid models combining audio and textrepresentations for improved accuracy.