The Event Semantic Modeling And Application Of Chinese Story

Natural Language Processing is an interdisciplinary research area at theborder between linguistics and artificial intelligence aiming at developingcomputer programs capable of human-like activities related to understanding orproducing texts or speech in a natural language. In recent years, with theincreasingly interesting in event semantic modeling for text, more and moreenterprises and academic organizations have used the resource descriptionframework (RDF) of World Wide Web Consortium (W3C) as the represtationspecification of the semantic knowledge base of text. The structured semanticreprestation of text will be benefit for the text understanding tasks such asinference and reasoning, and will be greatly improve the performance likequestion answering, information extraction, machine translation and othernatural language processing application.Regarding the text-driven comics automatic generation system as researchbackground and with the fact that event ontology model has strong semanticexpression and reasoning abilities, we propose a semantic ontology modelingmethod for story based on natural language processing technology. The goal ofour method is to obtain the semantic model from unstructured text, then usesemantic web techniques to realize interoperability machine, informationaccessing and automatic inference. The main work is summarized as follows:First of all, we analyze and summarize the research background and researchstatus of information extraction at home and abroad, especially the acquisition ofLexicalized Relation Resource; then we introduce the research background andresearch status of structured description of media data, as well as the currentsituation of event ontology model application in media data field.Next, we propose a semantic ontology modeling method for story. Thepileline consists of the following phases: general pretreatment method oriented tothe story visualization; n-ary shallow semantic frame elements extraction rules basedon open information extraction method; event types hierarchy suit for storyvisualization and trigger words combined with similarity method to recognizeevent types; event frame elements semi-automaticly mapping to event ontologywhich used as an event semantic annotation model for multimedia data(text, images, audio, sound, etc.).Finally, we apply our semantic ontology modeling method into thetext-driven comic automatic generation system. We first use the speechrecognition technology to obtain text description of pictures; then for both stories andtext description of pictures, we adopt the event semantic modeling method for story tochange the plain text into machine-readable semantic event stream format. Finally, weuse the query and reasoning mechanism of the semantic web to search thecorresponding images audio and other multimedia data information according tothe event stream of text.