Given a collection of graphs, find graphs in which a node labeled X is "near" a node labeled Y, for some definition of graph distance.
The motivation was to find documents with words near other words. Documents are normally considered strings of words, but in reality, language is not like that. At the very least, sentences are not strings but trees: one could have two words that are near each other in the parse tree but separated by a long interrupting phrase or clause. Pronouns allow concepts to be separated by many sentences, so make the connections a graph.
No comments :
Post a Comment