229.THE TARGET. THE SECOND (DENOTED AS P (¯ A| Q)) IS USED ¯D. BER...

222-229.

the target. The second (denoted as P (¯ a| q)) is used ¯

D. Bernhard and I. Gurevych. 2009. Combining lexical

to denote the translation probability with the answer

semantic resources with question & answer archivesfor translation-based answer finding. InProceedings

as the source and the question as the target. Table 7

of ACL, pages 728-736.

provides the comparison. From this Table, we see

P. F. Brown and V. J. D. Pietra and S. A. D. Pietra and

that the pooling strategy significantly outperforms

R. L. Mercer. 1993. The mathematics of statistical

the two baseline methods for question retrieval (row

machine translation: parameter estimation. Computa-

13 and row 14 vs. row 15).

tional Linguistics, 19(2):263-311.R. Bunescu and Y. Huang. 2010. Learning the relativeusefulness of questions in community QA. In Pro-

5 Conclusions and Future Work

ceedings of EMNLP, pages 97-107.X. Cao and G. Cong and B. Cui and C. S. Jensen. 2010.

In this paper, we propose a novel phrase-based trans-

A generalized framework of exploring category infor-

lation model for question retrieval. Compared to

mation for question retrieval in community question

the traditional word-based translation models, the

answer archives. InProceedings of WWW.

proposed approach is more effective in that it can

D. Chiang. 2005. A hierarchical phrase-based model for

capture contextual information instead of translating

statistical machine translation. InProceedings of ACL.

single words in isolation. Experiments conducted

H. Duan and Y. Cao and C. Y. Lin and Y. Yu. 2008.Searching questions by identifying questions topics

on real Q&A data demonstrate that the phrase-

and question focus. In Proceedings of ACL, pages

based translation model significantly outperforms