On filtering irrelevant results in peer-to-peer search
Yuval Merhav, Ophir Frieder · 2008
To improve Peer-to-Peer (P2P) search accuracy, we identify multiword entities in user queries using an entity corpus constructed based on n-gram statistics collected from LimeWire users. We show that by using a statistical function, we can build a reliable corpus and successfully parse each user query to its correct entities. Our hypothesis: With entity matching, overall precision can be significantly improved, while only slightly decreasing recall.