Comparative Analysis of Machine Learning and Deep Learning Models for Sentiment Analysis in Somali Language

Understanding and analysing sentiment in user-generated content has become crucial with the increasing use of social media and online platforms. However, sentiment analysis in less-resourced languages like Somali poses unique challenges. This paper presents the performance of three ML algorithms (DTC, RFC, XGB) and two DL models (CNN, LSTM) in accurately classifying sentiment in Somali text. The CC100-Somali dataset, comprising 78M monolingual Somali texts from the Common crawl snapshots, is utilized for training and evaluation. The study employed rigorous evaluation techniques, including train-test splits and cross-validation, to assess classification accuracy and performance metrics. The results demonstrated that DTC achieved the highest accuracy among ML algorithms, 87.94%, while LSTM achieved the highest accuracy among DL models, 88.58%. This study's findings contribute to sentiment analysis in less-resourced languages, specifically Somali, and provide valuable insights into the performance of ML and DL techniques. Moreover, the study highlights the potential of leveraging both ML and DL approaches to analyze sentiment in Somali text effectively. The results