Abstract

In the digital age, online communication influences how the public views brands, making sentiment analysis crucial for understanding consumer opinions and social trends. This study outlines the design and implementation of a FastText and custom-based sentiment analysis model specifically for the Nigerian linguistic context, which includes Yoruba, Igbo, Hausa, and Nigerian Pidgin. The system uses FastText embeddings for efficient classification and features a custom model fine-tuned for local language variations. It supports real-time and batch sentiment analysis of text files (TXT/PDF) and provides a visualization dashboard to show sentiment distribution. Additionally, it offers user authentication and history management for personalized use. Built with FastAPI, Flask, and Streamlit, the platform allows seamless interaction between backend processing and the user interface. Experimental results show high accuracy and usability across multilingual datasets, even with challenges like data imbalance and dialect diversity. Overall, this work contributes to Natural Language Processing (NLP) for African languages and provides a practical framework for brand monitoring and opinion mining in Nigeria’s changing digital landscape.

Keywords: Brand Monitoring, FastText, Natural Language Processing (NLP), Nigerian Languages, Sentiment Analysis.