Assessing relevance using automatically translated documents for cross-language information retrieval

摘要: This thesis focuses on the Relevance Feedback (RF) process, and scenario considered is that of a Portuguese-English Cross-Language Information Retrieval (CUR) system. CUR deals with retrieval documents in one natural language response to query expressed another language. RF an automatic process for reformulation. The idea behind it users are unlikely produce perfect queries, especially if given just attempt.The aims at improving queryspecification, which will lead more relevant being retrieved. method consists asking user analyse initial sample retrieved judge them relevance. In context, two main questions were posed. first relates user's ability assessing relevance texts foreign language, hand translated into their automatically second question concerns relationship between accuracy participant's judgements improvement achieved through process. In order answer those questions, this work performed experiment Portuguese speakers asked English documents, hand-translated Portuguese, Portuguese. results show machine translation as effective aiding assess relevance. In addition, impact misjudged documents performance overall moderate, varies greatly different topics. This advances existing research by considering carrying out experiments, aspects remained unexplored until now. contributions also include: investigation using new pair; design implementation stemming algorithm Portuguese; several experiments Latent Semantic Indexing contribute data points theory.