Analysis of answering questions using AI by categorization methods for text

Question Answering (QA) is a Computer Engineering area which consists of multi-disciplinary fields Artificial Intelligence (AI), Information Retrieval (IR), and Natural Language Processing (NLP). The main aim of these QA systems is to build systems that can answer questions asked by humans in a natural language according to the given passage. This process was challenging for earlier computers because of the hardware limitations and lack of software models needed to complete the tasks, which took a very long time to complete. Today, Computer Hardware advancements, especially in GPU units, made it possible to complete tasks in parallel much faster. Also, the recent improvements and research in AI models and software made it possible to use Pre-Trained models to achieve this goal much faster. In this thesis, one of the most popular models by Google, BERT (Bidirectional encoder representations from transformers), is Fine-Tuned, and the limitations are explored. A case study is made to understand how this Fine-Tuned model can help people in any area given. The results showed that working with large models and data sets still takes longer times for the training parts, and the Fine-Tuned Bert model performs better for the specific task it was designed.