Substitução de objetos em cache na internet usando modelo vetorial para comparação semântica da informação

AUTOR(ES)
DATA DE PUBLICAÇÃO

2005

RESUMO

This research work presents a new approach to compare objects stored in cache engines through semantic analysis of object contents. Semantic comparison between objects is used as a replacement strategy, differently from classical approaches that use objects physical characteristics as parameters to their selection and replacement strategies. A replacement strategy based on semantic comparison tends to keep in cache, objects that share semantics that are relevant to clients and consequently are related to each other. LSR (Least Semantically Related) was the first replacement strategy based on semantics. That strategy organizes objects in a tree structure of subjects. Each object is associated to one or more subjects from this semantic tree. However, LSR doesnt obtain semantics from objects by analyzing their contents. LSR assumes that each object is previously classified according to a taxonomy that is represented by a semantic tree. This research work presents a new strategy called LSR-VM (Least semantically Related Vector Model) that eliminates the need for a semantic tree of subjects. To establish the semantic comparison between objects inside the cache, LSR-VM uses a vector model strategy that is typically used to solve information retrieval problems. Vector model is a technique to organize, classify, compare and calculate the semantic distance between two objects. This research work presents the results obtained from LSR-VM strategy simulations, as well as comparisons to other traditional replacement strategies such as SIZE, LFU (Least Frequently Used) and LRU (Least Recently Used).

ASSUNTO(S)

ciencia da computacao memória cache internet sistemas de memória de computadores

ACESSO AO ARTIGO

Documentos Relacionados