Text Engineering Issues
- Development of the current Knowledge Base:
- The grammar takes advantage of the concept of levels of a grammar
- X-bar grammars
- Cascaded finite state automatas
- These levels are used for robust parsing so instead of
S -> word*, S -> Phrase*. These are automatically generated.
- This allows for a partitioning of the grammar.
- Agrammatical structure is absorbed as if it were ungrammatical.
- It runs with GATE.
- It has been used in the fifth and seventh Message Understanding
Competitions, so it does more than make parse trees.
- It's run in compiled common lisp
- The grammar could be compiled out (using eg. Lex) to run really
fast.
