Paper: Selective Sharing for Multilingual Dependency Parsing

ACL ID P12-1066
Title Selective Sharing for Multilingual Dependency Parsing
Venue Annual Meeting of the Association of Computational Linguistics
Session Main Conference
Year 2012
Authors

We present a novel algorithm for multilin- gual dependency parsing that uses annotations from a diverse set of source languages to parse a new unannotated language. Our motiva- tion is to broaden the advantages of multilin- gual learning to languages that exhibit signif- icant differences from existing resource-rich languages. The algorithm learns which as- pects of the source languages are relevant for the target language and ties model parame- ters accordingly. The model factorizes the process of generating a dependency tree into two steps: selection of syntactic dependents and their ordering. Being largely language- universal, the selection component is learned in a supervised fashion from all the training languages. In contrast, the ordering decisions are only influenced by languages w...