Invited Talk by Serge Sharoff
15 February 2017, by Reinhard Zierke
We are pleased to announce the invited talk of Serge Sharoff on Feb 15, 2017, immediately before our Stammtisch.
DATE: Feb 15, 2017
TIME: 17:30 s.t.
LOCATION: Von-Melle-Park 6 (Phil-Turm), Room Nr. 260
Visiting close relatives: Language Adaptation approach
ABSTRACT:
Many lesser-resourced languages are related to languages, which have better resources. For example, the Universal Dependency treebank has about 2 million words of training resources for Russian, and only 950 words for Ukrainian.
Similarly, the Autodesk Machine Translation dataset only covers three Slavonic languages, while only German is available among the Germanic languages.
To address such problems, I suggest a general approach, which can be called Language Adaptation, similarly to Domain Adaptation. In this approach, a model for a particular language processing task is built by lexical transfer of cognate words and by learning a new feature representation for a lesser-resourced language starting from a better-resourced one. In this talk I will demonstrate how language adaptation works in such training scenarios as Part-Of-Speech tagging, syntactic parsing and translation quality estimation. I'll also discuss two unsolved problems with this model: lexical transfer of MultiWord Expressions and word embeddings for homonymous words.
SHORT BIO:
Serge Sharoff, Associate Professor, Centre for Translation Studies, University of Leeds. Over the years he worked on a range of topics, covering Information Extraction, Natural Language Generation, Text Classification, tagging and parsing, as well as Machine Translation.