Skip to content

tmu-nlp/pmi-ppdb

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 
 
 

Repository files navigation

MIPA: Mutual Information Based Paraphrase Acquisition via Bilingual Pivoting

Our paraphrase acquisition method, called MIPA, first acquires lexical paraphrase pairs by bilingual pivoting and then reranks them by pointwise mutual information and distributional similarity. The complementary nature of information from bilingual corpora and from monolingual corpora makes our paraphrase acquisition method robust.

The English version of the paraphrase database was constructed from French-English bilingual corpus of Europarl-v7 and English monolingual corpus of English Gigaword 5th Edition. The Japanese version of the paraphrase database was constructed from English-Japanese bilingual corpora of The Kyoto Free Translation Task and Tanaka Corpus and Japanese monolingual corpus of BCCWJ.

The form of each line in the paraphrase database: original word <TAB> paraphrase <TAB> MIPA score

Use and/or redistribution of the paraphrase database is permitted under the conditions of Creative Commons Attribution-Share-Alike License 3.0.

For questions, please contact Tomoyuki Kajiwara at Tokyo Metropolitan University.

About

MIPA: Mutual Information Based Paraphrase Acquisition via Bilingual Pivoting

Resources

Stars

4 stars

Watchers

4 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors