Semantic Query Reformulation in Social PDMS

TL;DR

This paper introduces a relevance-based semantic query reformulation method for social peer-to-peer data management systems, improving accuracy and recall by selectively considering meaningful mappings and social links.

Contribution

It proposes a novel relevance notion and a reformulation approach that combines social links, semantic views, and gossiping techniques for efficient query rewriting in social PDMS.

Findings

Achieves greater recall than traditional methods

Enhances accuracy and flexibility in query reformulation

Utilizes social links and semantic views effectively

Abstract

We consider social peer-to-peer data management systems (PDMS), where each peer maintains both semantic mappings between its schema and some acquaintances, and social links with peer friends. In this context, reformulating a query from a peer's schema into other peer's schemas is a hard problem, as it may generate as many rewritings as the set of mappings from that peer to the outside and transitively on, by eventually traversing the entire network. However, not all the obtained rewritings are relevant to a given query. In this paper, we address this problem by inspecting semantic mappings and social links to find only relevant rewritings. We propose a new notion of 'relevance' of a query with respect to a mapping, and, based on this notion, a new semantic query reformulation approach for social PDMS, which achieves great accuracy and flexibility. To find rapidly the most interesting mappings, we combine several techniques: (i) social links are expressed as FOAF (Friend of a Friend) links to characterize peer's friendship and compact mapping summaries are used to obtain mapping descriptions; (ii) local semantic views are special views that contain information about external mappings; and (iii) gossiping techniques improve the search of relevant mappings. Our experimental evaluation, based on a prototype on top of PeerSim and a simulated network demonstrate that our solution yields greater recall, compared to traditional query translation approaches proposed in the literature.