Computational Linguistics Colloquium

Thursday, 12 January 2012, 16:15
Conference Room, Building C7 4

Supertagging-based HPSG parsing

Supertagging is a parsing technique for deep processing using large-scale lexicalized grammars. It has been shown to significantly improve the speed, and often the accuracy, of deep parsing. The task of supertagging is usually formulated as a sequence labeling task, in which each word in a sentence is tagged with a supertag that represents the syntactic/semantic characteristics of the word. Besides its importance for practical large-scale processing, we can see there interesting interactions between symbolic and statistical approaches to the syntactic disambiguation of natural language text.

In the talk, I will give an overview of the parsing techniques based on supertagging with some experimental results and show recent advances related to it. I'll also talk about the technique from the viewpoint of structured prediction in general, and explain the interaction between symbolic and statistical approaches in structured prediction.

If you would like to meet with the speaker, please contact Valia Kordoni.