[ Skip to the content ]

Institute of Formal and Applied Linguistics Wiki

[ Back to the navigation ]

You are here: start » seminar-20061023-k-hall

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

seminar-20061023-k-hall [2006/10/24 10:29]
 seminar-20061023-k-hall [2006/10/24 10:29] (current)
Line 1: Line 1:
 +====== Keith Hall: Multilingual Dependency Parsing ======
 +
 +
 +23. 10. 2006
 +Keith Hall (Center for Language and Speech Processing, Johns Hopkins
 +University, USA)
 +Multilingual Dependency Parsing and Applications
 +
 +[[ http://www.clsp.jhu.edu/~khall/talks/UFALSeminar.pdf | Slidy ]] [[ http://ufal.mff.cuni.cz/~bejcek/SFL/sfl_2006-10-23_Hall--dep_parsing.wma | audio ]]
 +
 +Abstract: Dependency parsing has recently come to the forefront of
 +interest in the statistical parsing community, culminating in the 2006
 +CoNLL shared task on multilingual dependency parsing. Many of the
 +competing teams made use of the Maximum Spanning Tree (MST) approach
 +pioneered by McDonald and Ribarov (McDonald et al. '05).
 +A disadvantage of the MST approach is that it requires structural scores
 +to be derived from parent-child links. This constrains the parsing models
 +to be based on very local structure; disallowing the explicit modeling of
 +subcategorization and valency as well as far simpler constraints (compound
 +adjectives, etc.).
 +In this talk, I present a two-stage dependency parser which combines a
 +K-best MST algorithm with a reranker. The advantage of such an approach is
 +that the model used by the reranker includes features defined over entire
 +tree structures. I present empirical results showing that "good" parses
 +appear in the first 50 hypotheses generated by the heavily constrained MST
 +models. Furthermore, I present reranking results that are competitive with
 +the state-of-the-art parsers. Results are presented for a subset of the
 +CoNLL competition languages as well as English.
 +Finally, I will introduce a framework for the application of dependency
 +parsing to tasks such as Speech Reconstruction and parsing of
 +resource-poor languages.
 +
 +
  

[ Back to the navigation ] [ Back to the content ]