Output details

A Transition-Based System for Joint Part-of-Speech Tagging and Labeled Non-Projective Dependency Parsing.

<22>The Empirical Methods in Natural Language Processing is one of the three top conferences in Computational-Linguistics. The paper describes the first successful approach that combines Part-of-Speech tagging and non-projective labeled dependency parsing. The system shows higher accuracy for both tagging and parsing. Furthermore, it provides a higher accuracy on text genres not included in the original training set, which is most critical for real world applications. The paper has lead to an invited talk (IMS,Stuttgart,Germany), and to a EU project proposal. Co-author Joakim Nivre was invited to Google-NY for a sabbatical and Google's translation system uses now a similar method.