Computational Linguistics, Machine Translation, Statistical Parsing, Statistical Semantics, Cognitive Modeling.
Natural Language Processing II (Statistical Structure in Language Processing
Profile Project AI-NLPL
We are interested in modeling natural language understanding for applications like machine translation, paraphrasing and summarization, tasks which we usually associate with human cognitive capability and expertise. Methodologically, we believe that best results can be achieved by devising advanced statistical learning techniques that fit with latent recursive sentence and document structure. Often we induce these latent recursive structures (graphs or simply trees) directly from data, but sometimes we have to define them following inspiration from Linguistic studies, which makes our field truly interdisciplinary. Some of our best known research lines that underly major industrial applications are data-oriented parsing; syntax-labeled phrase-based translation; alignment and permutation trees for word order translation models and their evaluation.