The Application Of The Planning Recognition In The Question And Answering System

With the prevalence of Internet, more and more people use the network to save andread data, which makes knowledge sharing realized. However, the difficulties faced to thissharing are how to inquire knowledge from the vast amounts of data quickly andaccurately. Combined traditional search engines with the application of natural languageunderstanding, the Q&A system can satisfy people’s needs.But, Q&A system’s problem model is essentially based on the context-free, withoutconsidering the correlation between questions. In General, we cannot guarantee thecomplement of user’s input. In that case, Q&A system cannot give search results basedon the information offered, but the user’s intention can be identified by the context of theprocess of contacting. Therefore, the paper broke the limit of the context-free, feedback tothe user appropriate results through the planning of the sentence. In addition, the questionwords of the sentence are usually near the words which express the purpose of theuser. Accurate in locating the question word of the questions can determine the typeof questions and help the retrieval.This is accomplished by parsing an individual sentence, completing the incompletesentences, replacing the pronoun and getting the implied information; and then identifyingthe question words in sentences, finding out the characteristics of the current sentence itemsand focus; finally, identifying the context of the sentence, using the recession-stylesemantics of the relationship between the current questions and previous questions,calculating the correlation, adding context to achieve a reference, in the same time avoidexaggerating the influence of questions. The focus of the sentence determines the scope ofthe search problems; the feature items of the sentence determine feedback informationcontent. In the answer extraction stage, use the knowledge map to express the structure ofthe article. The retrieval of knowledge use the planning theory, extracted from the focus onthe words as an observed action, based on observed actions and then narrows your search,finally locates the answer according to feature items.After analyzing questions in different areas, this paper proposes algorithm about thequestion-understanding. And has achieved the simulation of the process based on thefeasibility analysis. The answer extraction based on plan recognition has changed the previous answer extraction model, emphasized the accuracy and brief of the answer.Questions do not need to sieve a lot of connections from the information. In a word, thismodel improves the ability for the question answering system to understand naturallanguage intelligently.