汉语中方位参考点恢复研究

方位参考点恢复是自然语言空间语义理解中十分重要问题 .方位参考点恢复是在篇章中找方位词的参考点并补充上,得到完整的空间表达式 .目前,自然语言处理技术大多面向句子级,导致省略参考点空间表达式独立出现,使空间语义理解困难 .方位参考点恢复无疑可以解决类似问题 .在此提出基于有限知识的方位参考点恢复方法 .在句法分析基础上,以知网为常识库,结合有限知识识别空间表达式以及恢复方位参考点 .实验结果表明该方法比较令人满意 .

Research on the Resolution of the Landmark in Chinese

It is an essential problem to resolve the ellipsis of the reference in natural language processing. The resolution of the reference is to find the reference object, person or environment in the context in order to get a complete spatial expression and quantify the position and orientation expressed by the reference. In the present, the sentence including the spatial expressions with ellipsis of references is obtained based on many natural language processing technologies whose processing is in the level of sentence, which makes the following understanding in trouble. However, the resolution of references based on limited knowledge in this paper can resolve this problem. First of all, the spatial expressions are recognized based on Hownet and the linguistic knowledge after the text is parsed. Secondly, the candidate with the highest weight is proposed as the antecedent according to the limited knowledge including the part-of-speech, semantics and position. Finally, the experiments show that this method is satisfactory.