CAREER: Building Conceptual Natural Language Processing Systems for Practical Applications

本研究旨在开发一个概念自然语言处理（NLP）系统，该系统具有可适应的组件，可以很容易地针对不同的领域和应用进行定制。系统的架构由细粒度层组成，支持不同深度的文本处理。浅层支持语法处理，这可能足以完成一些信息检索任务，而深层支持语义和概念处理，以深入理解语言。该系统包括词性标注、介词短语附加、语义特征识别和概念提取等组件。每个组件都可以用最少的手工为新领域量身定制。分层架构还允许学生开发单独的组件，并将它们插入到更大的系统中进行实验。教育目标是将该系统作为面向年轻女孩的动手科学研讨会、面向高中生的暑期讲座、自然语言处理和机器学习的课堂项目以及研究生和本科生研究项目的基础。本研究的目的是开发自动或半自动地为新领域构建概念自然语言理解系统的技术。快速有效地生成概念句分析器是实现概念信息检索、文本分类和信息提取等许多实际应用的重要一步。

This research aims to develop a conceptual natural language processing (NLP) system with adaptable components that can be easily tailored for different domains and applications. The architecture of the system consists of fine-grained layers to support various depths of text processing. The shallow layers support syntactic processing, which may be sufficient for some information retrieval tasks, while the deeper layers support semantic and conceptual processing for in-depth language understanding. The system includes components for part-of-speech tagging, prepositional phrase attachment, semantic feature identification, and concept extraction. Each component can be tailored for new domains with minimal manual effort. The layered architecture also allows students to develop individual components and plug them in to the larger system for experimentation. The education goals are to use the system as the basis for a hands-on science workshop for young girls, for summer lectures to high school students, for class projects in natural language processing and machine learning, and for graduate and undergraduate research projects. The purpose of the research is to develop techniques for building conceptual natural language understanding systems automatically or semi-automatically for new domains. Generating conceptual sentence analyzers quickly and efficiently is an important step toward many practical applications, including conceptual information retrieval, text categorization, and information extraction.