What you have to know to use Leopar
Leopar is based on Interaction Grammars. The current version can be used with the French resources Frigram and can produce a very detailled syntactic analysis for complex French sentences. The fact that it is completely based on linguistic knowledge has several advantages:
- Overgeneration is very limited. Most of the ungrammatical French sentence are rejected by the Leopar/Frigram system.
- Leopar keeps ambiguity at every level of the analysis and it does not limit the number of syntactically correct solutions.
On the other hand, these choices have also some well known drawbacks:
- Leopar is not a robust parser. Our parsing algorithm cannot return partial parses and so, when a sentence failed to be parsed, nothing is returned.
- The fact that Leopar is a symbolic parser and that it does an almost exhaustive search for parse trees of the input sentence has two main consequences:
- It cannot parse very long sentences (a sensible size for input sentence is about 20 words);
- It may return a high number of solutions for medium size sentence.