Extraction of Entities and Relations调研

Extraction of Entities and Relations调研，后面将推出基于Stanford Parser 的具体实现

Phrase Dependency Parsing

•      A lot of product features are phrases

•      Phrase dependency parsing  extends traditional dependency parsing tophrase level

•      Relation extraction task can benefit from dependencies within a phrase.

The Approach

•      (1) Constructing phrase dependency  tree from results of chunking and dependency parsing;

•       (2)Extracting candidate product features and candidate opinion expressions;

•      (3) Extracting relations between product features and opinion expressions.

Phrase Dependency Tree

 

•      Toolkit: Stanford parser,MXPOST

Candidate Product Features and Opinion Expressions

•      Product Features

–  All NPs and VPs are selected as candidate product features.

–  Prepositional phrases (PPs) and adjectival phrases (ADJPs) are excluded.

•       Opinion Expressions

–  Use a dictionary which contains 8221opinion expressions to select candidates.

–  The tree distance between product featureand opinion expression in a relation should be less than 5.

RelationExtraction

•      Taking advantage of the kernel methods

•      Instances containing similar relationswill share similar substructures in their dependency trees

•      Generalize the definition by (Culotta andSorensen, 2004) to fit the phrase dependency tree.

Tree kernels fordependency trees

•       List of features assigned to nodes

 

Tree kernels for dependency trees

•       Matching function and similarity function

Kernel function for PDT

•      Add term Kin to handle theinternal nodes of a phrase

Relation Extraction Experiment

•       Corpus and features

•      Restricting nodes by m(ti, tj) is a way toprune the search space of matching subtrees

Relation Extraction Experiment

ILP and SRL Approach

Extract Entity and Opinion on the Forum and Twitter

•       Description of task:Who feels how on which aspects of which entity?

•        Sentiment towards involved entities

•        Sentiment towards participants

•        Extracting sentiment expressions from Twitter

 

Reference

•      Yejin Choi, Eric Breck, and Claire Cardie.2006. Joint extraction of entities and relations for opinion recognition. InProceedings of EMNLP.

•      Aron Culotta and Jeffrey Sorensen. 2004.Dependency tree kernels for relation extraction. In Proceedings of ACL 2004.

•      Yuanbin Wu, et al. 2009. Phrase Dependency Parsing for Opinion Mining.In Proceedings of EMNLP.