Kannada and Telugu Native Languages to English Cross Language Information Retrieval

摘要

One of the crucial challenges in cross lingualudinformation retrieval is the retrieval of relevant information foruda query expressed in as native language. While retrieval ofudrelevant documents is slightly easier, analysing the relevance ofudthe retrieved documents and the presentation of the results toudthe users are non-trivial tasks. To accomplish the above task,udwe present our Kannada English and Telugu English CLIRudsystems as part of Ad-Hoc Bilingual task. We take a queryudtranslation based approach using bi-lingual dictionaries. Whenuda query words not found in the dictionary then the words areudtransliterated using a simple rule based approach whichudutilizes the corpus to return the ‘k’ closest Englishudtransliterations of the given Kannada/Telugu word. Theudresulting multiple translation/transliteration choices for eachudquery word are disambiguated using an iterative page-rankudstyle algorithm which, based on term-term co-occurrenceudstatistics, produces the final translated query. Finally weudconduct experiments on these translated query using audKannada/Telugu document collection and a set of Englishudqueries to report the improvements, performance achieved forudeach task is to be presented and statistical analysis of theseudresults are given.