Extraction type machine intelligent reading understanding question-answering system

TL;DR: In this article, an extraction type machine intelligent reading understanding question-answering system is proposed, which aims to provide a question answering system capable of improving question answering query efficiency and practicability.

Abstract: The invention discloses an extraction type machine intelligent reading understanding question-answering system, and aims to provide a question-answering system capable of improving question-answeringquery efficiency and practicability. The method is realized through the following technical scheme: a document retrieval module constructs a full text search engine ES retrieval and semantic retrievaltwo-stage document retrieval system for massive text documents in a document library to form a preliminary document set of question and answer queries; the reading understanding module extracts deepsemantic features of questions and documents through a reading understanding pre-training model, judges the probability that answers exist in the documents by utilizing a multi-layer neural network model in combination with the semantic features and structural features, and achieves extraction of the answers by utilizing a pointer network; and the answer combination prediction module combines theanswers output by the reading understanding model to combine the redundant answers to obtain a possible answer list and a corresponding answer probability; and the model optimization module realizes training and optimization of a reading understanding model through a labeled document set, and provides a better reading understanding model for a question-answering system.