110 likes | 187 Vues
Sam Zhang. An Applied Ontological Approach to Computational Semantics. Ontology. - A web of knowledge organized in hierarchies. Odd One Out. Which of the following don't belong: Man Woman Tralfamadorian Boy Girl. Natural Language Toolkit. Python module Tutorial online Corpora.
E N D
Sam Zhang An Applied Ontological Approach to Computational Semantics
Ontology • - A web of knowledge organized in hierarchies
Odd One Out • Which of the following don't belong: • Man • Woman • Tralfamadorian • Boy • Girl
Natural Language Toolkit • Python module • Tutorial online • Corpora
Wordnet: Lexical relations • Synonym / Antonym • Hypernym / Hyponym : X and other Y • Meronym / Holonym • Entailments -- Verbnet
OpenCyc • Which of the following don't belong: • Washington • Jefferson • Lincoln • Palin
Algorithm • Distance/cosine measure from each word to other words • Word with longest average distance is odd one out
Heuristic • Two sided search through lexical relations • Dynamic word sense disambiguation -- Which part of speech? Which meaning?
Example • How do I find a common element between Toothbrush and Toilet? • from nltk.corpus import wordnet as wn • toothbrush = wn.synset("toothbrush.n.01") • toilet = wn.synset("toilet.n.01") • #algorithm to spider hypernyms until they match
Applications... And the Future • - Web 3.0 • - Literary criticism / Political analysis • ---------------------------------------- • - Supplement lexical relations from wordnet with own corpora search, especially from RSS feeds
Sources • Image: http://nltk.googlecode.com/svn/trunk/doc/book/ch02.html • Wordnet: http://wordnet.princeton.edu • NLTK: http://nltk.googlecode.com • OpenCyc: http://www.cyc.com/opencyc/overview • Ontological semantics, a history and overview: • http://crl.nmsu.edu/Staff.pages/Technical/sergei/book/realm.pdf