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.