alex@dutirt2.tudelft.nl (Alexander Vonk) (06/08/90)
Greetings everybody, Error handling in Natural Language Processing systems is one of the subjects of my master's thesis. Therefore, I'm very interested in (pointers to) literature, systems, etc. about this subject. Let me describe a little what I have in mind. Suppose you have a lexicon, a (context-free) grammar and a parsing method (I'm thinking about a chart parser or Blank's parsing method). You build a dialog system from it (adding some more fine technology) and want to incorporate some error handling to cope with real life language utterances. Error handling can be done in three stages: detection, analysis and correction. There are different kinds of errors. For example, let w=a(1)..a(n) be an input sentence to be parsed. The parser accepts w if w in L(G), with G the grammar used. But what happens if w is not in L(G), but w'=a(1)..a(k-1) a(k+1)..a(n) is in L(G). How can the parser restore a complete parse tree (or forest) from the incomplete parse? Now suppose another input contains an unknown or misspelled word; what can be done about it? What about partial sentences? I'll post a summary of the reactions I'll get. Thanks in advance, Alexander Vonk. +++ Alexander Vonk - Technical Univ. Delft, Netherlands +++ +++Phone: (NL) 015 - 78 64 12 (world) 31 15 78 64 12 +++ +++Mail: alex@dutirt2.tudelft.nl or alex@dutirt2.UUCP +++