Limits...
Identifying duplicate content using statistically improbable phrases.

Errami M, Sun Z, George AC, Long TC, Skinner MA, Wren JD, Garner HR - Bioinformatics (2010)

Bottom Line: For large-scale studies, submitting short phrases encased in direct quotes to search engines for exact matches would be optimal for both individual queries and programmatic interfaces.We have derived a method of analyzing statistically improbable phrases (SIPs) for assistance in identifying duplicate content.When applied to MEDLINE citations, this method substantially improves upon previous algorithms in the detection of duplication citations, yielding a precision and recall of 78.9% (versus 50.3% for eTBLAST) and 99.6% (versus 99.8% for eTBLAST), respectively.

View Article: PubMed Central - PubMed

Affiliation: Division of Translational Research, The University of Texas Southwestern Medical Center, 5323 Harry Hines Boulevard, Dallas, TX 75290-9185, USA. merrami@collin.edu

ABSTRACT

Motivation: Document similarity metrics such as PubMed's 'Find related articles' feature, which have been primarily used to identify studies with similar topics, can now also be used to detect duplicated or potentially plagiarized papers within literature reference databases. However, the CPU-intensive nature of document comparison has limited MEDLINE text similarity studies to the comparison of abstracts, which constitute only a small fraction of a publication's total text. Extending searches to include text archived by online search engines would drastically increase comparison ability. For large-scale studies, submitting short phrases encased in direct quotes to search engines for exact matches would be optimal for both individual queries and programmatic interfaces. We have derived a method of analyzing statistically improbable phrases (SIPs) for assistance in identifying duplicate content.

Results: When applied to MEDLINE citations, this method substantially improves upon previous algorithms in the detection of duplication citations, yielding a precision and recall of 78.9% (versus 50.3% for eTBLAST) and 99.6% (versus 99.8% for eTBLAST), respectively.

Availability: Similar citations identified by this work are freely accessible in the Déjà vu database, under the SIP discovery method category at http://dejavu.vbi.vt.edu/dejavu/.

Show MeSH
Four regions in the 2D space used for eTBLAST calibration to detect highly similar citations. Region B is the region in which eTBLAST predicts citations to be highly similar. Regions A and C do not contain many duplicate pairs of citations. Region D contains most MEDLINE citations and therefore most of the duplicate citations missed by eTBLAST. This figure is a modification of Figure 2 in Reference (14).
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License 1 - License 2
getmorefigures.php?uid=PMC2872002&req=5

Figure 1: Four regions in the 2D space used for eTBLAST calibration to detect highly similar citations. Region B is the region in which eTBLAST predicts citations to be highly similar. Regions A and C do not contain many duplicate pairs of citations. Region D contains most MEDLINE citations and therefore most of the duplicate citations missed by eTBLAST. This figure is a modification of Figure 2 in Reference (14).

Mentions: We have previously shown that eTBLAST can be used to detect highly similar citations (Errami et al., 2008). The calibration of eTBLAST for the detection of duplicate citations in MEDLINE has been described in detail (Errami et al., 2008). Briefly, when the title and abstract of a MEDLINE citation are queried against the MEDLINE database using eTBLAST, the algorithm returns a list of citations in order of their similarity to the query, as well as a similarity score for each. The most similar citation is, of course, the citation itself, labeled the Rank 1 citation. We label the most similar non-identical citation the Rank 2 citation, the second most similar non-identical citation Rank 3, etc. Figure 1 displays the similarity scores of Rank 2 citations plotted against the ratios of Rank 2 to Rank 1 similarity scores. The division of this plot into four distinct regions separates the citations pairs into groups with the following characteristics:Fig. 1.


Identifying duplicate content using statistically improbable phrases.

Errami M, Sun Z, George AC, Long TC, Skinner MA, Wren JD, Garner HR - Bioinformatics (2010)

Four regions in the 2D space used for eTBLAST calibration to detect highly similar citations. Region B is the region in which eTBLAST predicts citations to be highly similar. Regions A and C do not contain many duplicate pairs of citations. Region D contains most MEDLINE citations and therefore most of the duplicate citations missed by eTBLAST. This figure is a modification of Figure 2 in Reference (14).
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License 1 - License 2
Show All Figures
getmorefigures.php?uid=PMC2872002&req=5

Figure 1: Four regions in the 2D space used for eTBLAST calibration to detect highly similar citations. Region B is the region in which eTBLAST predicts citations to be highly similar. Regions A and C do not contain many duplicate pairs of citations. Region D contains most MEDLINE citations and therefore most of the duplicate citations missed by eTBLAST. This figure is a modification of Figure 2 in Reference (14).
Mentions: We have previously shown that eTBLAST can be used to detect highly similar citations (Errami et al., 2008). The calibration of eTBLAST for the detection of duplicate citations in MEDLINE has been described in detail (Errami et al., 2008). Briefly, when the title and abstract of a MEDLINE citation are queried against the MEDLINE database using eTBLAST, the algorithm returns a list of citations in order of their similarity to the query, as well as a similarity score for each. The most similar citation is, of course, the citation itself, labeled the Rank 1 citation. We label the most similar non-identical citation the Rank 2 citation, the second most similar non-identical citation Rank 3, etc. Figure 1 displays the similarity scores of Rank 2 citations plotted against the ratios of Rank 2 to Rank 1 similarity scores. The division of this plot into four distinct regions separates the citations pairs into groups with the following characteristics:Fig. 1.

Bottom Line: For large-scale studies, submitting short phrases encased in direct quotes to search engines for exact matches would be optimal for both individual queries and programmatic interfaces.We have derived a method of analyzing statistically improbable phrases (SIPs) for assistance in identifying duplicate content.When applied to MEDLINE citations, this method substantially improves upon previous algorithms in the detection of duplication citations, yielding a precision and recall of 78.9% (versus 50.3% for eTBLAST) and 99.6% (versus 99.8% for eTBLAST), respectively.

View Article: PubMed Central - PubMed

Affiliation: Division of Translational Research, The University of Texas Southwestern Medical Center, 5323 Harry Hines Boulevard, Dallas, TX 75290-9185, USA. merrami@collin.edu

ABSTRACT

Motivation: Document similarity metrics such as PubMed's 'Find related articles' feature, which have been primarily used to identify studies with similar topics, can now also be used to detect duplicated or potentially plagiarized papers within literature reference databases. However, the CPU-intensive nature of document comparison has limited MEDLINE text similarity studies to the comparison of abstracts, which constitute only a small fraction of a publication's total text. Extending searches to include text archived by online search engines would drastically increase comparison ability. For large-scale studies, submitting short phrases encased in direct quotes to search engines for exact matches would be optimal for both individual queries and programmatic interfaces. We have derived a method of analyzing statistically improbable phrases (SIPs) for assistance in identifying duplicate content.

Results: When applied to MEDLINE citations, this method substantially improves upon previous algorithms in the detection of duplication citations, yielding a precision and recall of 78.9% (versus 50.3% for eTBLAST) and 99.6% (versus 99.8% for eTBLAST), respectively.

Availability: Similar citations identified by this work are freely accessible in the Déjà vu database, under the SIP discovery method category at http://dejavu.vbi.vt.edu/dejavu/.

Show MeSH