Cross-Language Retrieval with the Twenty-One System
Wessel Kraaij, Djoerd Hiemstra · 1997
The EU project Twenty-One will support cross language queries in a multilingual document base. A prototype version of the Twenty-One system has been subjected to the Cross Language track tests in order to set baseline performances. The runs were based on query translation using dictionaries and corpus based disambiguation methods. 1 Introduction 1.1 Twenty-One project Twenty-One is a 2 MECU project with 11 partners 1 funded by the EU Telematics program, sector Information Engineering. The project subtitle is "Development of a Multimedia Information Transaction and Dissemination Tool". Twenty-One started early 1996 and is currently in its building phase. The Twenty-One database consists of documents in different languages, initially Dutch, English, French and German but extensions to other European languages are envisaged. The TREC Cross Language (CLIR) track task fits our needs to evaluate the system on the aspect of cross language retrieval performance. 1.2 TREC6 Although the de...