The Design and Implementation of Information Extraction Application for Stock Market Announcements in Chinese

摘要

This paper introduced the design and implementation of a Chinese IE Technology based stock market announcements information extraction system. The framework and deployment of the system were described. A simplified 3-tier IE model was discussed in detail. Among the 3 tiers, the algorithms used in template filling were mainly focused on. In this paper, the corpus-based statistical functionalism method is used for template filling. The Hidden Markov Model (HMM) and the IE algorithms based on it were introduced. The algorithm for model parameter acquisition was also analyzed and the Baum-Welch iteration algorithm was modified. HMM-based IE was implemented with the statistical data annotated manually from the domain texts.