KaggleX Final Project with Adejumobi Joshua | Kaggle

About this project: The Yoruba-RAG project focuses on enhancing the performance of large language models, like GPT-3, when handling questions in low-resource languages like Yoruba. The project involves web scraping from a Yoruba blog using Beautiful Soup, storing the data in a text file, and dividing it into smaller chunks. To effectively process Yoruba text, the Language-agnostic BERT Sentence Embedding (LaBSE) model is employed, and the results are stored in a Chroma database. This enriched database significantly improves GPT's ability …

blog data file gpt gpt-3 joshua kaggle language language models languages large language large language models low performance project questions rag scraping text web web scraping