UNSUPERVISED LEARNING OF DEEP PATTERNS FOR SEMANTIC PARSING

摘要

Using exemplary sentences, usage patterns and thematic roles ascribed in VerbNet to generate “deep pattern trees” for the exemplary sentences. Then, when an arbitrary natural language subject sentence is input, these deep pattern trees can be matched to the natural language subject sentence in order to assign thematic roles to at least some of the “grammatical portions” of the natural language subject sentence.

主权项

1. A method comprising:
receiving a set of exemplary sentences, including a first exemplary sentence, from a language net; parsing each of the exemplary sentences to yield a set of training constituency trees respectively corresponding to the exemplary sentences; performing maximal frequent subtree analysis on the set of training constituency trees to yield a set of at least one deep pattern trees including a first deep pattern tree corresponding to the first exemplary sentence; and assigning a first thematic role to a first node of the first deep pattern tree based on a thematic role assigned to a corresponding portion of the first exemplary sentence by the language net; wherein: each deep pattern tree is a constituency tree having a multiple layer hierarchy of nodes respectively corresponding to grammatical portions of a corresponding exemplary sentence from the language net; and at least the parsing step is performed by computer software running on computer hardware.