汉语中人称代词的消解研究

人称代词的消解是自然语言处理中十分重要的问题，人称代词消解，就是确定人称代词与先行语之间的相互关系，从而明确人称代词究竟指代什么对象，现有的许多应用系统，如文本摘要、信息抽取等采取了从文本中直接抽取句子的做法，而结果可能会含有某些无先行语的人称代词，使理解变得非常困难，人称代词消解无疑可以解决类似的问题。该文主要结合句类基本知识，根据人称代词所在语义块中的语义角色和人称代词对应的先行语可能的语义角色，给出了消解人称代词的基本规则。同时，作者也从句法的角度，结合局部焦点法给出了优选性规则。

Research on Chinese Pronominal Anaphora Resolution

Natural Language provides people with rich mechanisms for varying the expression of the same meaning. However. It also challenges computational researchers to do with a great number of nontrivial problems, one of them is anaphora resolution. In some application fields of discourse processing, such as Automatic Summarization, Information Retrieval, and Information Extraction etc., the final processing result is usually generated by directly extracting some representative sentences from text or document, in which, some pronouns could be contained without their antecedents. So, It is very important to resolve pronominal anaphora in Natural Language Processing(NLP), and that has attracted the attention of increasing researchers. Many approaches developed offer approximation solutions, including principle based such as purely syntactic ones to semantic and pragmatic ones, and statistics based ones. In this paper, sentences category(SC) based resolution tactics of Chinese pronominal anaphora is presented. In this method, the semantic role of pronoun and potential objects will be analyzed, and antecedent will be chosen from these objects according to the semantic relation between them and pronoun. SC is a semantic category of sentences within HNC(Hierarchical Network of Concept) Pattern. According to the HNC theory, every sentence can be mapped into one and only SC and the number of SC in Natural Language is determinate; Every SC consists of definite primary semantic chunk and indefinite secondary semantic chunk, and there exist conceptual constraints between primary semantic chunks in a sentence, as well as between primary semantic chunk and parts of secondary semantic chunks. In order to resolve anaphora, constraint rules and preference rules are given in this paper, which are related to the above conceptual constraint, and four kinds of anaphora relations will be discussed as follow: (1) in one semantic chunk; (2) between two primary semantic chunks; (3)primary semantic chunk and secondary semantic chunk; and (4) between two semantic chunks in different sentences.