Information Extraction as a Basis for Multi-Faceted Text Categorization Systems

该项目的目标是基于一种称为“信息提取”的自然语言处理技术，开发新的文本分类、文本分割和文本摘要算法。信息提取技术提供了基于单词的信息检索系统所不支持的语言分析级别，但它比深入的自然语言处理技术更健壮和可扩展。信息提取特别适合文本分类，因为许多分类问题需要识别关键字分析无法捕获的角色关系和上下文区别。该项目涉及构建一个多方面的文本分类系统，该系统支持多种文本处理功能，包括多类分类、主题分割和特定领域的文本摘要。目标是在多个文本语料库和不同的类别集上取得良好的性能。这项研究代表了一种新的文本分类方法，许多企业和政府机构都有关键的应用，包括文本路由和过滤等相关任务。

The goal of this project is to develop new algorithms for text categorization, text segmentation, and text summarization based upon a natural language processing technique called `information extraction`. Information extraction techniques provide a level of linguistic analysis that is not supported by word-based information retrieval systems, but are more robust and scalable than in-depth natural language processing techniques. Information extraction is particularly well-suited for text categorization because many categorization problems require the identification of role relationships and contextual distinctions that cannot be captured by keyword analysis. The project involves building a multi-faceted text categorization system that supports multiple text processing capabilities, including multi-class categorization, topic segmentation, and domain-specific text summarization. The objective is to achieve good performance on multiple text corpora and different category sets. This research represents a new approach to text categorization, for which many businesses and government agencies have critical applications including related tasks such as text routing and filtering.