A scalable cross-language metasearch architecture for multilingual information access on the Web

[Some figures poorly reproduced in original] MT Summit VII____________________ Sept. 1999 A Scalable Cross-Language Metasearch Architecture* for Multilingual Information Access on the Web Yoshihiko Hayashi, Genichiro Kikui, Toshiaki Iwadera NTT Cyberspace Laboratories 1-1 Hikari-no-Oka, Yokosuka, 239-0847. Japan E-mail: hayashi@nttnly.isl.ntt.co.jp Abstract This position paper for the special session on "Multilingual Information Access" com- prises of three parts. The first part re- views possible demands for Multilingual Information Access (hereafter, MLIA) on the Web, and examines required technical elements. Among those, we, in the sec- ond part, focus on Cross-Language Infor- mation Retrieval (hereafter, CLIR), par- ticularly a scalable architecture which en- ables CLIR in a number of language com- binations. Such a distributed architecture developed around XIRCH project (an in- ternational joint experimental project cur- rently involves NTT, KRDL, and KAIST) is then described in a certain detail. The final part discusses some NLP/MT related issues associated with such a CLIR archi- tecture. 1 Introduction A survey report titled "Web Languages Hit Parade" published in 1997 [Babel, 1997] estimated that more than 80% of the Web home pages are written in En- glish, followed by German, Japanese, French, and so on. While the estimation method described was rela- tively rough, the figure seems to be reasonable. What about the current and/or the future situa- tion? To our knowledge, there is no comparable in- formation source which gives the current estimation. An international online marketing company however makes another type of estimation, saying that the number of people on the net whose native language is other than English will overcome that of English- native people by Y2K [Global Reach, 1999]. As the Web is expected to grow continuously over the globe, * The XIRCH project described in this paper is joint work with MunKew Leong (KRDL, Singapore), and Key-Sun Choi (KAIST, Korea). this estimation would not be very misdirected. Nat- urally with this estimation, networked resources writ- ten in languages other than English are supposed to increase as well as demands for seeking those resources using a variety of languages. As discussed in the next section. CLIR definitely plays an essential role in MLIA. Introduced below is the "Grand Challenge" statement1 discussed in the wrap-up panel session of 1997 AAAI Spring Sympo- sium on Cross-Language Text and Speech Retrieval2 . Given a query in any medium and any lan- guage, select relevant items from a mul- tilingual multimedia collection which can be in any medium and any language, and present them in the style or order most likely to be useful to the querier, with iden- tical or near identical objects in different media or languages appropriately identi- fied. This is, as seen, a general and strong statement which enoughly covers many research issues on intelli- gent multimedia/multilingual information access tech- nichs/systems. Goals of MLIA are naturally implied by this statement. In the next section, we try to fig- ure out technical ingredients relevant to MLIA on the Web with this statement in mind. 2 Ingredients for MLIA on the Web The …

Content truncated to 500 words