Alignment of Spanish and English TREC Topic Descriptions

A technique is described for aligning TREC topic descriptions that is capable of producing a small multilingual test collection which can be used for cross language ad hoc and routing evaluations Meth ods for measuring the degree of degradation induced by the necessary approximations are described and illustrated using examples from an evaluation of two cross language routing techniques Although the experiments were conducted on a relatively small test collection using existing TREC relevance judg ments the results suggest that cross language routing is practical and that the investment required to produce a truly multilingual test collection for the TREC multilingual track would be justi ed