Turnitin: an improvement to similarity scoring

Turnitin have announced a change to the way similarity scores in Originality Reports are generated, when staff have chosen to exclude matching text in the Bibliography.

Details are provided in this announcement: An Improvement to Similarity Scoring.

The key passage is

When bibliographic material is excluded from an Originality Report, we will recalculate the Similarity Score based on the remaining content only. This update will provide you with a more accurate score.

 

Turnitin: exclude bibliography

 

The following example may help to demonstrate what the effect of these changes will be.

A student submits an essay where the total word count is 2500, including a bibliography of 500 words.

Turnitin generates an overall similarity score of 40% i.e. it finds matches for 1000 out of 2500 words.

Let’s assume that all 500 words in the bibliography produce a match.

With the old system, if we chose to ‘Exclude Bibliography’, Turnitin would perform a new calculation as follows:

matching words (total matching words – matching words in bibliography) / total words in document

= (1000 – 500) / 2500

=500/2500

= 20%.

 

With the new system the calculations would be as follows:

matching words (total matching words – matching words in bibliography) / words to be searched (total words in document – words in bibliography)

= (1000 – 500) / (2500 – 500)

= 500 / 2000

= 25%.

 

This does seem to be a more meaningful way of calculating the score. As ever, however, staff should treat the total similarity score with caution – it is necessary to look at the context of individual matches to form a judgment on whether plagiarism has taken place, or whether it’s simply an example of poor academic practice.