Jointparser

It performs syntactic parsing, shallow semantic parsing and predicate identification.
And it is one of the few parsers that simultaneously learns and annotates syntax and semantics.

We extended the Eisner algorithm to annotate semantics by assigning semantic links at each dependency scoring step. The learning is based on an averaged Perceptron.

For efficiency reasons, some syntax-based features used in the semantic classifier are pre-computed. The predicate identification is done as a previous step. The implementation of our model does not give us an optimal syntactic-semantic tree as we pre-compute some features.

The system description can be found at:

• Xavier Lluís and Stefan Bott and Lluís Màrquez A Second-Order Joint Eisner Model for Syntactic and Semantic Dependency Parsing In Proceedings of the CoNLL-2009 Shared Task bib.
• Xavier Lluís and Lluís Màrquez A Joint Model for Parsing Syntactic and Semantic Dependencies In Proceedings of the CoNLL-2008 Shared Task bib.
• Xavier Lluís, Advisor: Lluís Màrquez Joint Learning of Syntactic and Semantic Dependencies Master’s thesis, Technical University of Catalonia, 2008.

Software used in this demo:

• FreeLing
  POS tagger and lemmatizer
• whatswrong
  dependency structure visualizer

To try this parser just write a sentence: