摘要 |
<p>The present invention discloses an apparatus for determining the similarity between paraphrase identification-based sentences, and a method for operating same. The present invention comprises: a phrase analysis step for extracting predicate argument tuples corresponding to a first sentence and a second sentence, respectively, by designating a correlation between two or more words included in each phrase in a sentence, on the basis of a grammatical relationship between the words comprising the sentence; a similarity measuring step of measuring the similarity between one or more predicate argument tuples extracted from the first sentence and one or more predicate argument tuples extracted from the second sentence; and an approximate alignment step of carrying out an approximate alignment for measuring the similarity between the first sentence and the second sentence, on the basis of the similarity of the predicate argument tuples that is measured, thereby enabling an approach to a paraphrase identification problem with a supervised learning-based automatic categorization model.</p> |
申请人 |
KOREA INSTITUTE OF SCIENCE & TECHNOLOGY INFORMATION;CHOI, SUNG PIL;CHUN, HONG WOO;JEONG, CHANG HOO;CHOI, YUN SOO;SONG, SA KWANG;JUNG, HAN MIN |
发明人 |
CHOI, SUNG PIL;CHUN, HONG WOO;JEONG, CHANG HOO;CHOI, YUN SOO;SONG, SA KWANG;JUNG, HAN MIN |