Research On Technology Of Intelligent Question Answering Based On Service Planning And Data Prediction

Mainstream Web Question Answering(WQA)systems extracts the answer of a question by selecting one page from numerous Web pages,and to rapidly locate and display contents required by users.It has already been extensively applied to various fields.However,for many complex problems,we need the answers to be generated by two or more Web application dynamically,rather than acquired from a single page.With the prevalence of Web services,more and more information is dynamically produced in the form of Web services,which makes it possible to rapidly extract effective information from mass information in the Internet on the base of Web service.However,for complex problems in real life,the functions that can be provided by a single Web service are limited.Therefore,there is a need to seek answers on the base of multiple information sources,to integrate several Web services,and even to utilize intelligent computational models like data prediction.Based on this thought,this thesis proposes an intelligent Web Question Answering(WQA)system based on Web service planning and data prediction,with tourism field as the background.It uses the user interface of natural language to answer questions,and to give plans and suggestions.Moreover,instructions of users are executed by combining a group of Web services.In question answering,the intelligent WQA system may predict the variation trend of relevant information according to the context,and provide the best suggestions for users dynamically and adaptively according to the future evolution trend,so as to help users make decisions.This thesis involves two technologies:the first one is WQA technology based on Web service planning,and the second one is nonlinear multivariate data prediction technique blended in WQA technology.(1)WQA techniques based on Web service planningIn mainstream WQA,natural language is quite suitable for text matching as text input.But in WQA based on Web services,need to address the issue of how natural language corresponding to the required services.At the same time,WQA based on service planning needs to solve the problem of searching and planning multiple information sources to answer users' questions.Regarding the issue of involving Web services in WQA,this thesis proposes a QAtheory model based on dynamic service planning.Regarding the issue of building the problem domain and planning domain of HTN in QA theory model,this thesis proposes mapping model of question and services: Formal description for the verbs and objects of natural language is completed by utilizing Logical Grammar,question and services matching is completed by using semantic distance algorithm based on OWL-S description of Web services.In the simulation experiment of natural language question analysis,services matching and planning,the natural language calling service realized by the technique proposed in this thesis has gained a good result.(2)Modeling techniques for multivariate data prediction model to realize predictive question answeringThe future tour arrangements can be planned more reasonably by integrating data prediction into WQA in the tourism field.Relevant data of tourism(such as weather)have characteristics of multiple dimensions,non-linearity and instability.It is complex to rapidly establish a set of prediction models according to the historical data.With these concerns,this thesis carries out a research on time series of nonlinear mathematical modeling regarding the issue that WQA cannot conduct predictive answering:As for the issue of parameter estimation in the prediction model,multi-component parallel parameter estimation scheme and damping parameter algorithm are proposed.Regarding the issue of error correction in prediction model,this thesis introduces non-linear evolution technology.The optimal error function is extracted by utilizing its powerful automatic searching function and combining with evolution pattern of model error in the historical data.On the basis of the above work,regarding the modeling problem of the prediction model of WQA in the tourism field,this thesis studies a method of evolutionary modeling for one-dimensional and multi-dimensional prediction equations,and proposes a new algorithm of adaptive value based on vector difference and a new thought of secondary evolution.The numerical experiments show that the proposed method can adaptively mine its potential kinetic equation according to evolution characters of the historical data,and makes the model have a better forecasting ability within a certain period of time.