Key Technique Of Chinese Spoken Open-Domain Question Answer System

The Natural Language Question Answer technology is in the natural language processing domain an extremely popular research direction. It synthesized has utilized each kind of natural language processing technology. The natural language question and answer man-machine contact surface, is precise and real-time is the natural language interrogator-responder system three big research and development goal. Among them, the accuracy is the Chinese Question Answer System priority target. In order to achieve this goal, in the user interrogative sentence processing aspect, must carry on the correct participle, the synonym to the user input interrogative sentence expands, the famous entity sign note, the syntax analysis, answer type sign note and so on processing. Regarding knowledge source documents, also must carry on similar processing. Regarding said based on the traditional IR technology Question Answer System, but also needs a comprehensive survey user interrogative sentence and the text piece between similarity computational method. It can be said, the realization interrogator-responder system needs the technology involves computational linguistics the aspects.Chinese Spoken Language Interactive Open-Domain Question Answering System is precisely a front research direction which emerges in this foundation, This article briefed Chinese Spoken Language Interactive Open-Domain Question Answering System development present situation and the commonly used essential technology. Chinese Spoken Language Interactive Open-Domain Question Answering System including four main parts: Pronunciation Processing, the Question Analysis and Processing, the Information Retrieval, the Answer Extract with the answer choice. This article introduced separately these four main constituent involves essential technology, Put forward the standardized question type storehouse concept and the simple model, Finally also introduced the Chinese Spoken Language Interactive Open-Domain Question Answering System simple realization and the appraisal question.Pronunciation Processing synthesizes two parts including the speech recognition and the Pronunciation. The user use natural language spoken language inquiry question first by the speech recognition partially through the pretreatment, the characteristic withdraws with the pattern recognition forms the system to think the best recognition output, transforms the spoken language question into the text question, by supplies the following module analysis and processing. But the pronunciation synthesis partial main functions are transform the system production text answer through the spoken language synthesis technology for the spoken language answer feed back in the user. This system feeds back first after the speech recognition the text answer in the user gain user's opinion realization spoken language interaction.The Question Analysis Processing part is carries on the analysis and processing after the speech recognition question. Chinese Spoken Language Interactive Open-Domain Question Answering System first carries on the question the analysis work, this process analysis effect has the important influence to the behind treating processes. Below the problem analysis are partial needs to complete several parts of work: Carries on key word, based on question factor and so on type to the question which the participle as well as the lexical category sign note, the determination question type, withdraws has problems to the key word carries on the suitable expansion. In the article we mainly introduced the question classification, And put forward the standardized question type storehouse concept.The Information Retrieval partial duties are search the related documents the key words which withdraws with front to the documents storehouse in. The information retrieval module returns is some most related documents. May transfer directly in this series information retrieval module had the retrieval system also to be possible to transfer on Internet search engine for instance Google. This article also designed an Frequently-Asked Question Database. After passes through problem analysis processing the question, can automatically commonly used Frequently-Asked Question Database seeks candidate answer set, through computation sentence similarity, the answer which matches returns for the user. Also can automatically renew and maintain the Frequently-Asked Question Database. Does not request each interrogator-responder system in the TREC (Text REtrieval Conference) all to have to have own information retrieval module, because the TREC can provide the most related 1,000 documents for each question.General search engine like Google returns is pile of homepages, but Chinese spoken language interactive opening territory interrogator-responder system needs to return is the brief answer. Thus, the related documents which searches through the information retrieval module must submit to the answer extracts the module to refine the answer. The answer may have several kind of types, possibly is a speech, or is several speeches, also possibly is several words or the phrase. Regarding these asked the time place the question, may use the very short sentence to reply, but regarding the inquiry reason, the event question needs the long sentence to be able to reply. Therefore the answer extracts also needs to determine the answer based on the question type the type.This article elaborated had been considered was the next generation INERNET main application method, namely spoken language interactive Open-Domain Question-Ansering system. In domestic and foreign proposed for the first time the spoken language operates, the standardization alternately asked the question bank the development and applies two new methods. These two methods thorough research and the use has the value extremely.