Intelligent Question Answering System Sentence Similarity Computing Applications

In the intelligent answering, machine translation, intelligent retrieval and other fields, are among the key sentence similarity computing technology. Answering system can be calculated through the sentence similarity is found in the Frequently. Asked Questions database and user to match the queries raised by the answer to return to relatively more accurate answer. Based on the research process, and then click on the original meaning, words, sentences of three levels of similarity computation have been studied, which analyzed the sentence similarity calculation. On this basis, this thesis propose a wide range of sentence similarity calculation method, and experimental analysis prove the effectiveness of the method, the final design of the entrance area of a wide range of sentences answering system and applied to the similarity calculation them.This thesis is mainly reflected in the following areas:1. For the current advantages and disadvantages of word similarity calculation methods are described, and its improved calculation of sentence similarity to pave the way below.2. Quantified sentence from a multidimensional perspective of various types of information to build a wide range of sentence similarity computing model. Utilization of the model sentence length, the same number of key words, sentences, semantic, keywords, sentences and other information, using the weighted sum method to determine the similarity between sentences.3. This thesis uses genetic algorithm to determine the characteristics of a variety of weight between the sentences, to ensure the accuracy and reliability of the weights.4. Designed a recruitment area answering system (EAQS, Enrollment Area Question-Answering System) model, the wide range of sentence similarity calculation method is applied to one of the frequently asked questions database (FAQ) .The questions match, and gives the sentence similarity in the FAQ library implementation process.