DMCA
A Bayesian Approach to Unsupervised Semantic Role Induction
Cached
Download Links
| Citations: | 16 - 4 self |
BibTeX
@MISC{Titov_abayesian,
author = {Ivan Titov and Alexandre Klementiev},
title = {A Bayesian Approach to Unsupervised Semantic Role Induction},
year = {}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
We introduce two Bayesian models for unsupervised semantic role labeling (SRL) task. The models treat SRL as clustering of syntactic signatures of arguments with clusters corresponding to semantic roles. The first model induces these clusterings independently for each predicate, exploiting the Chinese Restaurant Process (CRP) as a prior. In a more refined hierarchical model, we inject the intuition that the clusterings are similar across different predicates, even though they are not necessarily identical. This intuition is encoded as a distance-dependent CRP with a distance between two syntactic signatures indicating how likely they are to correspond to a single semantic role. These distances are automatically induced within the model and shared across predicates. Both models achieve state-of-the-art results when evaluated on PropBank, with the coupled model consistently outperforming the factored counterpart in all experimental set-ups. 1
Keyphrases
bayesian approach semantic role induction syntactic signature factored counterpart different predicate bayesian model experimental set-ups distance-dependent crp semantic role coupled model chinese restaurant process first model induces unsupervised semantic role labeling refined hierarchical model state-of-the-art result single semantic role
