We've carried out a comprehensive quality assessment of our content tagging system.
Both usability and accuracy was assessed relative to the leading 3rd party system (Refinitiv® Intelligent Tagging) in a test with a random sample of 200 news & press release articles tagged with both the Refinitiv® and JotAI systems. We did not
Accuracy was defined as presence of tag(s) best describing main article subject.
.PNG)
.PNG)
Both systems have produced a comparable number of usable results.
Refinitiv's tags tended to be more specific and fewer, whereas JotAI's tend to be broader and more in number, e.g. 'COVID-19 in the United Kingdom' versus 'COVID-19, United Kingdom'. JotAI produced more false positives, but these tended to be very obscure subjects that would be unlikely to ever be searched for, so their overall impact was negligible.
Overall, the results have exceeded our expectations. Our next goal is an even more accurate iteration that would cover more industry-specific domains and reduce the number of false positives.