Context-Aware Deep Learning Approach for Answering Questions

Abstract

Training neural networks to read and comprehend natural language documents still poses a great challenge. In the recent past, large scale training and test data sets have been made available for testing machine reading systems based on their ability to answer unseen questions, given the context. In Question-Answering, the model generates answers from the given context for the questions (given as input. Various machine learning methods can be used to build such systems. For understanding natural language, the model should be able to convert the sentences or paragraphs into a representation that is internal to the model (understandable by it), to be able to generate valid answers. Valid answers are the ones which answer the question asked correctly.[1] This work attempts to design a deep learning model to read and comprehend the context provided and provide answers to the posed questions in natural language accurately with very little knowledge of language structure.