There were lots of methods for anaphora resolution. As described in Dr.Wang Houfeng's survey paper, there were some theoretical methods such as, Hobbs Algorothm 1997, Center Theory of Grosz & Sidner 1983. But they had not been used directly.
The classical techniques of this topic were two main aspects.
The first was based on syntax. There were so many method on it, such as RAP of Lappin & Leass in 1994, Kennedy & Boguraev in 1996, Mitkov in 1998.
The another was based on corpora. For example, there were some methods like Simple Co-appear, statistical method, Design tree, clustering and Memory based Learning.
When I read this survey paper, I bethought of that based on the syntax parser we can make good anaphora resolution. But when I asked Mjs. He told me that right now, we can not realize the syntax component analysis. If I give a sentence to the parser, I can not get the subject, object, and so on.
It seems difficult of this topic. I should read more papers on it.
没有评论:
发表评论