Research For Event Extraction Method In Specific Domain Based On Tree Conditional Random Field

Event extraction is referred to extract events occurred in the non-structural text. The extracted events should contain structural attributes such as time, location, object and content of the event. As a research branch of information extraction, event extraction has become a research hot spot along with the emergency of vertical search engine and been applied to many related domains like information retrieval, vertical search engine.Aims at huge amounts of Curriculum Vitae (hereby referred as CV) documents scattered on the Internet, the research work of the thesis focus on how to automatically recognize and download the CV documents on the Internet by text classification, and, how to extract structural events from downloaded CV documents with unstructured text based on conditional random field.A summarization and review about methods of text classification and models of event extraction is presented. The paper also compares and analyzes progress of text classification and event extraction, their methods and models.Based on social tagging and document vector model, a social tagging and text frequency combined text vector model is proposed. Combining the new model with the traditional text classification approaches, the precision of classification can improved.According to the hierarchal property of resume information, tree conditional random field is used to model the hierarchal information of resume text, and a prototype system is completed.Aiming proposed model and event extraction algorithm based on tree conditional random field, many experiments of text classification and event extraction are implemented. The results prove the efficiency of our methods.