Uso de paralelismo para aprimorar o software de busca por indícios de plágio Miss Marple

Abstract

The developed work had as objective the improvement of Miss Marple plagiarism checking tool, by using parallel-distributed processing techniques. Firstly, it was identified that tool, in its original version, used to take long time for processing documents. This fact limits its usage and applicability to a small set of documents. Preliminary tests had shown that tool takes few advantage of current multi-core architectures, suggesting the need of developing a new version of the tool, so named Parallel Miss Marple. This new version of the tool is able of better extracting the power of current parallel and distributed computing architectures. For taking advantage of multi-CPU architectures, the new tool makes use of thread programming paradigm. For multicomputer distribution, it is used the Java RMI API. The new tool has been tested by submitting a set of digital texts documents, where the processing times of the original and the parallelized versions of the application where measured by using a Profiler tool. Tests show a satisfactory increase in performance by using the new version of the program.