The Calculation Method Of User Advisory Text Similarity Research

With the development of the Internet technology and the rhythm of human life became more faster, traditional acquisition knowledge based on books, journals, newspapers and other paper-based information carrier can not meet people’s needs. In contrast, today’s more a slogan------ spread problem, look for "Du niang", that is, people gradually shift the way to get information to the network up. Faced with a vast network of resources,people how to quickly and accurately find the answer you want, which is all the experts and scholars research priorities.How to make your computer to correctly understand the issues raised by users, and how quickly and accurately find the answer to the question of resources in the vast amounts of data? This is launching an important research topic------ text similarity calculations. Many experts and scholars have proposed lots of themselves models and methods appropriate text similarity calculation, greatly promoted the development and application of natural language understanding disciplines. Similarity computing applications is very broad, common are: business intelligence consulting systems, information retrieval systems, machine translation systems, large data processing systems, etc. Based on previous research, the semantic similarity calculation method by users advisory text in more in-depth research, the main work is as follows:First, I have carefully study there is existing computational model for user consultation and calculation methods sentence similarity in short text, and learning theory techniques about people advisory text similarity calculation. Analysis of the research status and methods of user consultation short text similarity, and summarizes the advantages and disadvantages of commonly used model in its field of application.Secondly, On the basis of existed semantic similarity about user advisory text representation model, Proposed a abstract-knowledge modle to calculate users advisory text semantic similarity. Use the abstract knowledge model, we can split the user consultation text sentences problem into the "Key-word" and "the abstract knowledge" two parts. Thus the user consulting on text similarity computing split into the "key-word" and "abstract knowledge" similarity calculation.Finally, on the study of existed user advisory text semantic similarity calculation algorithm, gives an improved user consultation short text similarity algorithms. Using the "key-word" and "abstract knowledge" semantic similarity calculation represent that user advisory text calculate semantic similarity, this paper introduced the KNN algorithm, so clever puts the "vote electoral law" applied to the text advisory sentence similarity algorithms. Experimental results show that the proposed model abstract knowledge of the accuracy of text similarity computing has greatly improved role in the results closer to the artificial result of the judgment.