Research On The Construction Of News Text Event Chain Based On Predicate - Argument Structure

With the arrival of the big data era, people look increasingly out of their depth in searching for the information they need in the vast sea of information. It is imperative to research the Chinese Information Extraction Technology. News text is one of the key information which people always focus on and the updates of news online can be counted as thousands everyday. How to get faster and more comprehensive event information at a minimum cost that gets more and more attention. As a superior concept of lexical chain, event chain construction technology can get the key information of text effectively and make up for the disadvantage that lexical chain has bigger sentences recognition particle size and easy to be jammed by noise so that it can improve the accuracy and completeness of information extraction. Event chain construction technology can be as a basic technology of information retrieval and automatic abstracting. It has widely application value in many aspects such as text classification, text mining and Chinese semantic library construction.The article is mainly studied from the following four aspects and gets some achievements:(1) Text pretreatment technology was briefly introduced and the related approaches of segmentation and part of speech tagging was be studied and compared. Then we introduced the method of building the lexical chain, which is lower seat concept of event chain. We used the words similarity calculation based on wiki category-link as the basis of lexical chain construction.(2) We researched the method of building Predicate-Argument Expression in this study. The expression, which is a kind of expression mechanism based on the structure of predicate and argument, is also the basic unit of the event chain. In the forth chapter of this article, we determined the definition and rule of Predicate-Argument Expression by analyzing the rules of linguistics and event semantics and selected the characteristics based on semantic dependency relations to achieve the construction of Predicate-Argument Expression.(3) We researched the method of building the passage event chain in this study. To make the Predicate-Argument Expression as object and lexical chain as foundation, we constructed the event chain that can cover the whole information of passage. In the fifth chapter of this article, we firstly optimized lexical chain to make it easy to use. Then we described the model of event chain and the definition rules were given. Finally we got the concrete method to build the passage event chain.(4) Using the above results and the news text as specific subjects, the event chain structure prototype system was designed and completed.