Melange: Components for Cross-Lingual Retrieval

We present the finalized version of our cross-lingual search engine Melange, and results obtained by running it on WebCLEF topics in an attempt to solve Mixed Monolingual and Multilingual tasks. We concentrate on certain features of the system which are relevant to the CLIR field and which can be developed further independently. These are our data extraction and indexing methods, our language detection module (with an accuracy of 88% on WebCLEF query strings), PageRank ranking scheme and query translation.