Massively parallel acceleration of a document-similarity classifier to detect web attacks
Journal of Parallel and Distributed Computing
| Hi-index | 0.00 |
This paper proposes a hardware approach to improving information retrieval efficiency on the vector space model. We design a heterogeneous multicore system, within which auxiliary retrieval-oriented Intellectual Properties (IP) cores perform term counting as basic operations. Moreover, a memory system is also designed to supply data efficiently. Since term counting is a highly frequent operation ininformation retrieval, we hope the overall efficiency will improve significantly as a result of hardware implementation of term counting. The experiment shows that our system has a speedup of 3 to 7 for different input data. We have also analyzed the influence of data patterns on performance.