Ingenious: Text Summarization and Question Answering

Abstract

Deep learning has attained many remarkable advancements over the past few years and is now rapidly developing in the NLP field. Abstractive automatic text summarization indicates the process of employing computer software that summarizes a document without altering the actual intent of the content. Several use cases for automatic summarization include creating headlines, summarizing scientific documents, segmenting search results, and summarizing product reviews. In the period of the Information explosion, large amounts of data, and the Internet, the ability to express the core meaning of information concisely will help address the information overload crisis. Traditional Techniques often relied on extractive summarization, which involves selecting and rearranging existing sentences or phrases from the source text to create a summary that may lack coherence and fail to generate novel sentences. Deciding what information to include or exclude and how to compress the content without losing important details is challenging. In this research, a BART-based model is used (a denoising autoencoder for pre-training inter-sequence models) to develop a data set trained for automatic analysis and summarization of long texts and articles. The model is pre-trained in English. Additionally, it uses a model based on the trained dataset to answer questions from the text. Our approach is based on textual answers provided by the community, making it easier to implement and answer more complex questions. The proposed schema explores various techniques such as question-and-answer classification and query generation. Finally, the test results are attached and evaluated.