Evaluation of Natural Language Processing on Local and Public Cloud Resources

概要

論文の詳細を見る
In high-performance computing, parameter survey applications, which consists of a lot of independent tasks with different input parameters are typical and conventionally run on local computing resources with batch scheduler systems such as SGE and Condor. Our targeted PSA application is a Natural Language Processing (NLP) application. It uses Support Vector Machine (SVM) to learn and eliminate abuse words from the commercial bulletin board systems (BBSs) documents. Security issues become important because it processes confidential documents including personal informations. In this paper, we implemented an effective and secure hybrid mechanism with our previous work, InterS. It adds IaaS Cloud resources to the batch schedulers when the local resources are insufficient due to failures or overhead by external application tasks. It utilizes Amazon VPC service to enable the secure computation of our NLP application. InterS run the NLP application in our experiment and the evaluation of time efficiency is shown from the results.
2010-07-27