Abstract:
"
The research study is conducted on intrinsic plagiarism detection in Sinhala language
documents. There are considerably low number of studies done on the plagiarism
detection and authorship verification for Sinhala language. This research proposes an
anomaly detection-based approach classify text portions based on anomalous behavior
when compared to the neighboring context for the featured extracted using word
embedding based approach. In the study multiple feature extraction methods and
anomaly detection algorithms and supervised algorithm were used to conduct a series
of experiments to identify the combination which perform best for the Sinhala
languages. Study uses paragraph level features to distinguish segments with anomalistic
behavior. Proposed solution was able classify plagiarized content with an accuracy of
85% with a f1-score of 0.40."
