Answers Generation Based on English Textual Analyzer(AGETA)

Abstract

Accurate and contextually appropriate answer generation is increasingly important for applications in virtual assistants, educational tools, search engines, and so on, for answering any question about information found across electronic libraries. This research, “Answers Generation based on English Textual Analyzer (AGETA),” hopes to generate the correct answer as a complete comprehension sentence. It receives a passage with a related list of questions in English as unstructured typed texts. It performs the English textual analysis process using a series of natural language processing (NLP) techniques, followed by a hybrid method that combines extractive techniques with linguistic analysis to build an effective answer generator, as a set of its main applied techniques: tokenization, part-of-speech tagging, cosine similarity, T5 model, and then applying grammar-checking mechanism and English syntactic rules. AGETA was tested on questions related to different passages, with performance measured using two types of accuracy measurements. The first type is human performance, which achieved 92% for short answers and 96% for expanded answers. The second type is Sentence Transformers (BERT-based models), which achieved 90%. This indicates that the generated answers exhibit considerable unity with the corresponding ground truth answers. The proposed approach has potential applications in education, research, and customer support by enhancing the accessibility and relevance of textual information.