Using WordNet-based Context Vectors to Estimate the Semantic Relatedness of Concepts Siddharth Patwardhan University of Utah Ted Pederseb University of Minnesota, Duluth 1 Monday, July 29, 13 

a WordNet-based measure of semantic relatedness by combining the structure and content of WordNet with co-occurrence information derived from raw text 2 2 Monday, July 29, 13 

First Order Co-occurrence words occur near each other in a corpus of text ex: 1st (police, car) 1st (White House, Obama) 3 3 Monday, July 29, 13 

Second Order Co-occurrences 1st (w_1, w_2) 1st (w_3, w_2) ----------------------- 2nd (w_1, w_3) ex: 1st (police, car) 1st (mechanic, car) ----------------------- 2nd (police, mechanic) 4 4 Monday, July 29, 13 

Second Order Context Vector - Gloss Vector 1. Create a Word Spaceɿ a co-occurrence matrix where each row can be viewed as a first order context vector 2. Create a Gloss Vectorɿ Treat the dictionary definition of a concept as a context, and finding the resultant of the first order context vectors of the words in the definition. The gloss vector if formed by adding the vectors. word w_1 w_2 ... w_n w_1 4 3 ... 0 w_2 5 4 0 ... w_n 0 1 3 lamp: an artificial source of visible illumination fork: cutlery used to serve and eat food corpus dictionary 5 5 Monday, July 29, 13 

fork: cutlery used to serve and eat food The gloss vector of fork if formed by adding the first order context vectors of cutlery, serve, eat, and food. 6 6 Monday, July 29, 13