Academia.eduAcademia.edu

Figure 5 – uploaded by Paolo Rosso

See full PDFdownloadDownload figure
Fig. 5 Illustration of the cross-language explicit semantic analysis model

Figure 5 Illustration of the cross-language explicit semantic analysis model

Related Figures (9)

Fig. 1 Taxonomy of text plagiarism types, along with approaches to detect them (Meyer zu Eissen et al.
Fig. 1 Taxonomy of text plagiarism types, along with approaches to detect them (Meyer zu Eissen et al.
Fig. 2 Retrieval process of cross-language plagiarism detection, inspired by Stein et al. (2007) At first sight this process may appear rather generic, but the underlying con- siderations become obvious when taking the view of the practitioner: since plagiarists make use of the World Wide Web, a plagiarism detection solution has to use the entire indexed part of the Web as reference collection D’. This requires the retrieval of candidate documents D’, with ID’, < |D'|, since a comparison of d, against each Web document is infeasible. The following sections discuss particularities of step |
Fig. 2 Retrieval process of cross-language plagiarism detection, inspired by Stein et al. (2007) At first sight this process may appear rather generic, but the underlying con- siderations become obvious when taking the view of the practitioner: since plagiarists make use of the World Wide Web, a plagiarism detection solution has to use the entire indexed part of the Web as reference collection D’. This requires the retrieval of candidate documents D’, with ID’, < |D'|, since a comparison of d, against each Web document is infeasible. The following sections discuss particularities of step |
Fig. 3. Retrieval process of the heuristic retrieval step for cross-language plagiarism detection
Fig. 3. Retrieval process of the heuristic retrieval step for cross-language plagiarism detection
expressions from L to L’. Models of the third and fourth kind have to be trained on an aligned corpus that contains documents from the languages to be compared. The two approaches differ with respect to the required degree of alignment: comparable alignment refers to documents in different languages, which describe roughly the same topic, while parallel alignment refers to documents that are translations of each other and whose words or sentences have been mapped manually or heuristically to their respective translations. Obviously the latter poses a much higher requirement than the former. The following models have been proposed:
expressions from L to L’. Models of the third and fourth kind have to be trained on an aligned corpus that contains documents from the languages to be compared. The two approaches differ with respect to the required degree of alignment: comparable alignment refers to documents in different languages, which describe roughly the same topic, while parallel alignment refers to documents that are translations of each other and whose words or sentences have been mapped manually or heuristically to their respective translations. Obviously the latter poses a much higher requirement than the former. The following models have been proposed:
Table 1 Estimated length factors for the language pairs L-L’, measured in characters.
Table 1 Estimated length factors for the language pairs L-L’, measured in characters.
Fig. 6 Length model distributions that quantify the likelihood whether the length of the translation of d into the considered languages is larger than ldl. In this example, d is an English document of 30,000 characters (vertical line), corresponding to 6,600 words A value of j > 1 implies Idl < Id'l for d and its translation d’
Fig. 6 Length model distributions that quantify the likelihood whether the length of the translation of d into the considered languages is larger than ldl. In this example, d is an English document of 30,000 characters (vertical line), corresponding to 6,600 words A value of j > 1 implies Idl < Id'l for d and its translation d’
Table 2 Results of Experiment | for the cross-language retrieval models
Table 2 Results of Experiment | for the cross-language retrieval models
Table 3 Results of Experiment 2 for the cross-language retrieval models
Table 3 Results of Experiment 2 for the cross-language retrieval models
Table 4 Results of Experiment 3 for the cross-language retrieval models
Table 4 Results of Experiment 3 for the cross-language retrieval models
Related topics:
Cognitive SciencePlagiarism DetectionLanguage Resources
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved