Please use this identifier to cite or link to this item: http://ir.lib.seu.ac.lk/handle/123456789/3125
Title: Accelerating text-based plagiarism detection using GPUs
Authors: Jiffriya, M.A.C.
Jahan, M.A.C. Akmal
Gamaarachchi, Hasindu
Ragel, Roshan G.
Keywords: CPU
GPU
NVIDIA
CUDA
Jaccard similarity
Vector space model
Hashing strategy
Thread
Block
Issue Date: 20-Jul-2016
Publisher: IEEE
Citation: 10th International Conference on Industrial and Information Systems. 18th-20th Dec, 2015. Peradeniya, Sri Lanka.
Abstract: Plagiarism is known as an unauthorized use of other’s contents in writing and ideas in thinking without proper acknowledgment. There are several tools implemented for textbased plagiarism detection using various methods and techniques. However, these tools become inefficient while handling a large number of datasets due to the process of plagiarism detection which comprises of a lot of computational tasks and large memory requirement. Therefore, when we deal with a large number of datasets, there should be a way to accelerate the process by applying acceleration techniques to optimize the plagiarism detection. In response to this, we have developed a parallel algorithm using Computer Unified Device Architecture (CUDA) and tested it on a Graphical Processing Unit (GPU) platform. An equivalent algorithm is run on CPU platform as well. From the comparison of the results, CPU shows better performance when the number and the size of the documents are small. Meantime, GPU is an effective and efficient platform when handling a large number of documents and high in data size due to the increase in the amount of parallelism. It was found out that for our dataset, the performance of the algorithm on the GPU platform is approximately 6x faster than CPU. Thus, introducing GPU based optimization algorithm to the plagiarism detection gives a real solution while handling a large number of data for inter-document plagiarism detection.
URI: http://ir.lib.seu.ac.lk/handle/123456789/3125
https://doi.org/10.1109/ICIINFS.2015.7399044
Appears in Collections:Research Articles

Files in This Item:
There are no files associated with this item.


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.