Whats the difference between relevance, confidence and score? (from TRIT Webinar Feb 8)
Whats the difference between relevance, confidence and score?
Using news as an example, Relevance: How relevant is a company to the news story. A news story can mention sometimes 5 to 10 companies. But mostly the News story is about one or two companies. Those main companies will get high relevance score (80%).
In some cases there are market coverage stories where many companies are mentioned, each company has some coverage in a single paragraph. The story is not about a particular company, but each paragraph has some valuable info on a company. In this case relevance of 50% will be given to companies.
Relevance 20% is low - i.e. the company is mentioned but with minimal importance to the story - the story is not about this company in particular.
Relevance score can be 0% (for example Reuters is mentioned in a News story because it is the News provider. But the story has nothing to do with Thomson Reuters. Same with Rating agencies that are quoted providing rating score. Same with Social Media companies like Facebook when mentioned “share on facebook” – the story has nothing to do with Facebook).
“Continuous relevance” is a more granular version of the relevance score (more granular than 0, 20, 50, 80%).
Confidence: How confident Intelligent Tagging is that, the actual entity is of the correct type
Score: This is the confidence in matching to the Organization Authority (OA). Intelligent Tagging tries to find a company name in OA . The higher the Score, the higher the probability that Intelligent Tagging matched the correct company record and found the correct PermID for that company.
This question was first asked in How to Enhance Your Search Platforms webinar (8th of Feb). Please see here: How to Enhance Your Search Platforms