| At present time,how to make computers understand human natural language and achieve human-machine dialogue applying simulation of the natural language of human language communication process have been an important field of artificial intelligence—Natural Language Processing.QA system is a much broader prospect of research in natural language processing areas.And its research also triggered a boom in related areas on the otherreading comprehension(RC).RC system can analyze a given story in natural language automatically and return an answer for the given question by using the information of the given document.At present,there are many researches about English reading comprehension.On the contrary,the researches of Chinese reading comprehension are scarce.In this paper,we try to investigate some technologies for Chinese QARC.In this paper,we will give you the detailed introduction about the CRCC v1.1(Chinese Reading Comprehension Corpus) built by Shanxi University. This corpus provides the resource for the technologies research.Then we will describe word-level features and 25 syntax-level features,and we construct the rules against time,human,location,number types on RC system.In this experiment,the rules are constructed based on the relationship between questions and candidate answers sentence.And last,we will carry an experiment for four categories of Chinese questions applying rules based on CRCC,at the same time we select the rules' weight applying orthogonal design method.A QA system based on rules is built,and the performance of the system is evaluated though the experiments.In this dissertation,we adopt the HumSent accuracy to evaluate the system performance.The result shows that manual prepared by the rules is basically to get more satisfactory results.The experiments of the four types QA on CRCC get 88.24%,80%,94.12%and 89.61%HumSent separately.But the rules are constructed on human knowledge,which is subjective and short of objective. |
PDF Full Text Request