Keyword search considering user's preference in P2P networks

Abstract

When a user searches data in a P2P network, it is general to use a keyword-search, which acquires data items whose keywords match keywords the user specified. In a naive keyword-search strategy, however, users usually get a large number of data items regardless whether they actually need these data items or not because this strategy returns data items without considering the user's preference. In this paper, we propose a novel search strategy considering user's preference in P2P networks. In our strategy, each peer predicts the user's preference based on keywords featured in the data items. When searching, a peer appends some additional keywords representing the user's preference to the query in addition to the keyword that the user specifies. Peers which receive the query calculate the similarity between their holding data items and the information on the user's preference appended by the query sender. As a result, if the similarity is high, the peer returns its holding data items whose keywords match the received query. Thus, users can avoid getting unnecessary data items. In addition, each peer learns the user's preference using keywords featured in data items that the user accesses.