TY  - JOUR
T1  - Dynamic Caching for Semantic Oriented Adaptive Search in Unstructured P2P Networks
AU - Ramachandran, G. AU - Selvakumar, K. 
JO  - Asian Journal of Information Technology
VL  - 13
IS  - 3
SP  - 138
EP  - 149
PY  - 2014
DA  - 2001/08/19
SN  - 1682-3915
DO  - ajit.2014.138.149
UR  - https://makhillpublications.co/view-article.php?doi=ajit.2014.138.149
KW  - Peer to peer search
KW  -semantic search
KW  -vector space model
KW  -latent semantic index
KW  -dynamic caching
AB  - Most of the existing unstructured Peer to Peer (P2P) supports 
  the filename or keyword based limited search techniques. This study proposes 
  a novel Semantic Oriented Adaptive Search (SOAS) strategy based on semantic 
  content. It is a multi-layer architecture model which utilize a dynamic caching 
  technique to achieve effective search on unstructured P2P network. The proposed 
  scheme constructs as a two-tier P2P network with ultra peers of high connectivity 
  based on a power law model. The novelty of the proposed scheme is that the query 
  is processed through multi-tier summary indexing framework. The proposed approach 
  extensively used Vector Space Model (VSM) and Latent Semantic Index (LSI) to 
  derive local indices from summarized semantic vectors. Query searching concedes 
  through a round of searches using derived indices from semantic data objects. 
  If one search fails, the next round search is invoked sequentially in the order 
  of local index, cache index, response index, global index and adaptive search 
  among ultra peers. The proposed SOAS produces a high success rate and generates 
  a minimal amount of network traffic for an effective content search. Dynamic 
  Time to Live (TTL) based cache consistency is proposed where each ultra peer 
  dynamically caches the responses of previously requested queries based on the 
  query popularity rate and TTL. An implementation and large scale simulation 
  are performed to evaluate the proposed approach. The experimental result proves 
  that the proposed system performs better than the existing approach in terms 
  of accuracy, response time, success rate and cache hit.
ER  -