Towards Deeper MT: Parallel Treebanks, Entity Linking, and Linguistic Evaluation

Ankit Srivastava, Vivien Macketanz, Aljoscha Burchardt, Eleftherios Avramidis

In: Proceedings of The Workshop on Deep Language Processing for Quality Machine Translation. Workshop on Deep Language Processing for Quality Machine Translation (DeepLP4QMT) befindet sich AIMSA 2016 September 10 Varna Bulgaria ISBN 978-619-7320-03-9 Institute of Information and Communication Technologies Bulgarian Academy of Sciences 2016.


In this paper we investigate techniques to enrich Statisti- cal Machine Translation (SMT) with automatic deep linguistic tools and evaluate with a deeper manual linguistic analysis. Using English–German IT-domain translation as a case-study, we exploit parallel treebanks for syntax-aware phrase extraction and interface with Linked Open Data (LOD) for extracting named entity translations in a post decoding frame- work. We conclude with linguistic phenomena-driven human evaluation of our forays into enhancing the syntactic and semantic constraints on a phrase-based SMT system


deeperMT.pdf (pdf, 243 KB )

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence