Sentiment Analysis On Social Media In Afaan Oromo Using Multilingual Bert

Abstract

. Sentiment Analysis And Opinion Mining Have Become Crucial In Today's Digital Age As People Increasingly Share Their Thoughts And Emotions On Various Online Platforms. With The Development Of Powerful Language Models And Advancements In Speech Technology, Sentiment Analysis Models Have Been Trained On Diverse Languages And Datasets. However, African Languages, Including Ethiopian Languages Like Afaan Oromo, Have Often Been Overlooked In Sentiment Analysis Research Due To Their Linguistic Diversity And Limited Available Resources. In This Study, We Aimed To Address This Gap By Conducting Sentiment Classification On Afaan Oromo Datasets Using The State-Of-The-Art Multilingual Bidirectional Encoder Representation From Transformers (Mbert) Model. We Collected A Comprehensive Dataset Consisting Of 6742 Posts, Tweets, And Comments From Different Social Media Platforms. Each Entry In The Dataset Was Manually Annotated As Either Positive Or Negative Sentiment. By Fine-Tuning Multilingual Bidirectional Encoder Representation From Transformers (Mbert) Model On Our Afaan Oromo Dataset, We Achieved An Impressive Accuracy Of 85.96%. This Demonstrates The Effectiveness Of The Mbert Model In Capturing Sentiment In Afaan Oromo Language And Highlights The Importance Of Considering Linguistic Diversity In Sentiment Analysis Research. Our Findings Contribute To The Growing Body Of Work Aimed At Expanding Sentiment Analysis To Underrepresented Languages, Enabling A More Inclusive Understanding Of Online Opinions And Emotions.