a) b) c) Improving Quality of Pivot Translation by Context in Service Coordination Yohei MURAKAMI a), Rie TANAKA b),andtoruishida c) Web 1. Web 26.8% 30.9% 21.3% 21% 1 n n(n 1) Department of Social Informatics, Kyoto University, Kyotoshi, 606 8501 Japan C&C C&C Innovation Initiative, NEC Corporation, Ikoma-shi, 630 0101 Japan a) E-mail: yohei@i.kyoto-u.ac.jp b) E-mail: r-tanaka@ak.jp.nec.com c) E-mail: ishida@i.kyoto-u.ac.jp [1] Web 1 Internet World Stats 2011 5 http://www.internetworldstats.com/stats7.htm D Vol. J97 D No. 1 pp. 165 172 c 2014 165
2014/1 Vol. J97 D No. 1 2 [2] 2. 2. 1 [3] 1 (a) paper (a) (b) (c) 1 Fig. 1 Issues in coordinating translation services. (b) feast feast feast (c) Kranich crane 2. 2 SMT SMT [4], [5] Linguistic Annotation Language 166
[6] Linguistic Annotation 3. Fig. 2 2 Context-based translation coordination. 2 2 3 Fig. 3 Multi-agent architecture for pivot translation. 3 1 Bilingual word pairs 4. EuroWordNet [7] ILI Interlingual Index 167
2014/1 Vol. J97 D No. 1 [8] 4 2 3 2 1( ) 5 feast banquet Festessen( ) {, feast, Festessen} (a) Loop type (b) Transition type 4 Fig. 4 Lexical triangle. 5 Fig. 5 Loop triangle representing sense of gochisou. [9] [10] [11] 5. n CA MTA i 1 i n 6 7 CA s 1 s 1 w 1 n n-tuple T 1 8 T 1 s 1 T 1 MTA 1 10 i MTA i s i T i MTA i s i+1 s i o i m i+1 Q i 11 i +1 T i+1 12 T i i +1 s i+1 n-tuple Algorithm 2 T i s i MTA i 7 Algorithm 3 MT i 168
6 Fig. 6 CA Algorithm of coordinator agent CA. GET-WORD-PAIRS-USED-BY-MT o i c i+1 P i 6 7 Algorithm 4 CREATE-WORD-PAIRS-TO-BE-USED T i P i T i T i c i+1 8 c i+1 T i i o i n-tuple i +1 o i m i+1 Q i 10 11 T i i i +1 n-tuple m i+1 MTA i T i m i+1 MODIFY-TRANSLATED-SENTENCE P i Q i s i+1 7 MTA i Fig. 7 Algorithm of translation agent MTA i. Algorithm 3 10 m i+1 [12] 2( 1(c) ) 8 1(c) n-tuple 8 CA s 1= Kraniche kommen im Herbst nach Japan. 169
2014/1 Vol. J97 D No. 1 MTA 2 s 2 s 3= P 2 {{crane, }} T 2 crane Q 2 MTA 2 MODIFY- TRANSLATED-SENTENCE s 3= s 3 CA 6. 8 Fig. 8 Example of coordinating german-english and English-Japanese translation. s 1 T 1 s 1 T 1 MTA 1 MTA 1 MT 1 s 1 t 1= Cranescomeintheautumn in Japan. GET-WORD-PAIRS- USED-BY-MT P 1 P 1 {{Kranich, crane}} MTA 1 CREATE-WORD-PAIRS- TO-BE-USED Kranich crane T 1 T 1 Q 1=P 1={{Kranich, crane}} s 2=t 1= Cranes come in the autumn in Japan. MTA 1 s 2 Q 1 CA s 2 Q 1 CA SELECT- POSSIBLE-N-TUPLES T 2 {Kranich, crane, } T 2 s 2 T 2 MTA 2 6. 1 30,000 15,627 13,757 21,914 21,914 Web5 500 [13] 58% 40% 6000 50% 38% 6. 2 5. [14] A B C A B C 5 (5: All, 4: Most, 3: Much, 2: Little, 1: None) 170
(a) 3 Much 5 All (a) 5-grade evaluation (b) 3 Much 5 All 2 10 Fig. 10 Examples of improvement. (b) Relative evaluation 9 Fig. 9 Evaluation of German-Japanese pivot translation. 3 A 378 51 3 8 40 40 2.783 2.908 0.125 9 (a) B C 3 B 1 3 8.3 C 7.7 t 5 9 (b) C B -1 C 1 0 B 7 C 25 8 1% C 5 10 Kranich Parteien 7 171
2014/1 Vol. J97 D No. 1 7. Web 58% 40% S 24220002 24 28 [1] T. Ishida, Language grid: An infrastructure for intercultural collaboration, SAINT-06, pp.96 100, 2006. [2] K. Decker, K. Sycara, and M. Williamson, Middleagents for the internet, IJCAI-97, pp.578 583, 1997. [3] N. Yamashita and T. Ishida, Effects of machine translation on collaborative work, CSCW-06, pp.515 523, 2006. [4] M. Utiyama and H. Isahara, A comparison of pivot methods for phrase-based statistical machine translation, HLT-NAACL, pp.484 491, 2007. [5] H. Wu and H. Wang, Pivot language approach for phrase-based statistical machine translation, ACL 07, pp.856 863, 2007. [6] H. Kanayama and H. Watanabe, Multilingual translation via annotated hub language, MT-Summit IX, pp.202 207, 2003. [7] P. Vossen, EuroWordNet: A Multilingual Database with Lexical Semantic Networks, TDordrecht, Kluwer, Netherlands, 1998. [8] vol.6, no.2, pp.228 235, 1991. [9] Y. Wu, F. Li, R. Tanaka, and T. Ishida, Automatic creation of n-lingual synonymous word sets, SKG- 08, pp.141 148, 2008. [10] K. Tanaka and K. Umemura, Construction of a bilingual dictionary intermediated by a third language, COLING-94, pp.293 303, 1994. [11] Mausam, S. Soderland, O. Etzioni, D.S. Weld, K. Reiter, M. Skinner, M. Sammer, and J. Bilmes, Panlingual lexical translation via probabilistic inference, Artif. Intell., vol.174, no.9-10, pp.619 637, 2010. [12] J. Matsuno and T. Ishida, Constraint optimization approach to context based word selection, IJCAI-11, pp.1846 1851, 2011. [13] D. Kawahara and S. Kurohashi, Case frame compilation from the web using high-performance computing, LREC-06, pp.1344 1347, 2006. [14] vol.9, no.4, pp.569 579, 1994. 25 3 28 8 6 2006 2008 C&C 1976 1978 IEEE 172