Abstract: 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.