Linguistic Question-Answering Reasoning Based on Intelligent Perception of Attribute Weight

Abstract

Intelligent Question Answering System aims to realize the communication between human and machine. Based on the real data of question-answer, of front end design, and of daily chat in Chinese, this paper proposes a design method of Intelligent Question-Answering System based on deep neural network and optimizes it. Through real data from authentic view and online daily chat, the paper constructs the property weight, laying foundation for the system to play its role in the real scene such as text cleaning and word segmentation, part of speech tagging, word vector representation, word weight adjustment, etc. By using the mixed model of BOW and Skip-gram to represent the word vector, the paper describes the correlation between words and makes for the shortcomings of BOW. Meanwhile, it retains Bow’s excellent ability of discrete-feature processing.