Users often have vague or imprecise ideas when searching the e-commerce Web databases such as used cars databases, houses databases etc. and may not be able to formulate queries that accurately express their query intentions. They also would like to obtain the relevant information that meets their needs and preferences closely. In this paper, we present a new approach - QRR (query relaxation and ranking), for relaxing the initial query over e-commerce Web databases in order to provide relevant answer to the user. QRR relaxes the query criteria by adding the most similar values into each query criterion range specified by the initial query, and then the relevant answers which satisfy the relaxed queries could be retrieved. For relevant query results, QRR speculates the importance of each attribute based on the user initial query and assigns the score of each attribute value according to its "desirableness" to the user, and then the relevant answers are ranked according to their satisfaction degree to the user's needs and preferences. Experimental results demonstrate that QRR can effectively recommend the relevant information to the user and have a high ranking quality as well.
展开▼
