To plagiarise is to robe credit of another person's work. Particularly, plagiarism in text means including text fragments (and even an entire document) from an author without giving him the correspondent credit. In this work we describe our first attempt to detect plagiarised segments in a text employing statistical Language Models (LMs) and perplexity. The preliminary experiments, carried out on two specialised and literary corpora (including original, part-of-speech and stemmed versions), show that perplexity of a text segment, given a Language Model calculated over an author text, is a relevant feature in plagiarism detection.
Towards the exploitation of statistical language models for plagiarism detection with reference / Barron-Cedeno A.; Rosso P.. - ELETTRONICO. - 377:(2008), pp. 15-19. (Intervento presentato al convegno 18th European Conference on Artificial Intelligence, ECAI 2008 - Workshop on Uncovering Plagiarism, Authorship and Social Software Misuse, PAN 2008 tenutosi a Patras, grc nel 2008).
Towards the exploitation of statistical language models for plagiarism detection with reference
Barron-Cedeno A.;
2008
Abstract
To plagiarise is to robe credit of another person's work. Particularly, plagiarism in text means including text fragments (and even an entire document) from an author without giving him the correspondent credit. In this work we describe our first attempt to detect plagiarised segments in a text employing statistical Language Models (LMs) and perplexity. The preliminary experiments, carried out on two specialised and literary corpora (including original, part-of-speech and stemmed versions), show that perplexity of a text segment, given a Language Model calculated over an author text, is a relevant feature in plagiarism detection.File | Dimensione | Formato | |
---|---|---|---|
paper2.pdf
accesso aperto
Tipo:
Versione (PDF) editoriale
Licenza:
Licenza per Accesso Aperto. Creative Commons Attribuzione (CCBY)
Dimensione
157.58 kB
Formato
Adobe PDF
|
157.58 kB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.