Paper: A Joint Syntactic and Semantic Dependency Parsing System based on Maximum Entropy Models

ACL ID W09-1217
Title A Joint Syntactic and Semantic Dependency Parsing System based on Maximum Entropy Models
Venue International Conference on Computational Natural Language Learning
Session shared task
Year 2009
Authors

A joint syntactic and semantic dependency parsing system submitted to the CoNLL-2009 shared task is presented in this paper. The system is composed of three components: a syntactic dependency parser, a predicate clas- sifier and a semantic parser. The first-order MSTParser is used as our syntactic depen- dency pasrser. Projective and non-projective MSTParsers are compared with each other on seven languages. Predicate classification and semantic parsing are both recognized as clas- sification problem, and the Maximum Entropy Models are used for them in our system. For semanticparsingandpredicateclassifying, we focus on finding optimized features on multi- ple languages. The average Macro F1 Score of our system is 73.97 for joint task in closed challenge.