Title:Accelerating and Evaluation of Syntactic Parsing in Natural Language Question Answering Systems

Abstract: With the development of Natural Language Processing (NLP), more and more systems want to adopt NLP in User Interface Module to process user input, in order to communicate with user in a natural way. However, this raises a speed problem. That is, if NLP module can not process sentences in durable time delay, users will never use the system. As a result, systems which are strict with processing time, such as dialogue systems, web search systems, automatic customer service systems, especially real-time systems, have to abandon NLP module in order to get a faster system response. This paper aims to solve the speed problem. In this paper, at first, the construction of a syntactic parser which is based on corpus machine learning and statistics model is introduced, and then a speed problem analysis is performed on the parser and its algorithms. Based on the analysis, two accelerating methods, Compressed POS Set and Syntactic Patterns Pruning, are proposed, which can effectively improve the time efficiency of parsing in NLP module. To evaluate different parameters in the accelerating algorithms, two new factors, PT and RT, are introduced and explained in detail. Experiments are also completed to prove and test these methods, which will surely contribute to the application of NLP.