Detecting Cross-lingual Semantic Similarity Using Parallel PropBanks.

Abstract

This paper suggests a method for detecting cross-lingual semantic similarity using par-allel PropBanks. We begin by improving word alignments for verb predicates gener-ated by GIZA++ by using information avail-able in parallel PropBanks. We applied the Kuhn-Munkres method to measure predicate-argument matching and improved verb predi-cate alignments by an F-score of 12.6%. Us-ing the enhanced word alignments we checked the set of target verbs aligned to a specific source verb for semantic consistency. For a set of English verbs aligned to a Chinese verb, we checked if the English verbs belong to the same semantic class using an existing lexi-cal database, WordNet. For a set of Chinese verbs aligned to an English verb we manually checked semantic similarity between the Chi-nese verbs within a set. Our results show that the verb sets we generated have a high correla-tion with semantic classes. This could poten-tially lead to an automatic technique for gen-erating semantic classes for verbs. 1