Notions of Correctness when Evaluating Protein Name Taggers

by Olsson et al.

This paper introduces six notions of correctness to evaluate the performance of protein name taggers:

• sloppy - which means that the proposed hit contains parts of the correct solution (as well as other parts)
• protein name parts (pnp) - the hit is a part of the full protein name
• strict - equals a verbatim match
• left/right/left or right - the hit matches exactly on the left/right/left or right boundary

The authors suggest that notions different from the verbatim (strict) match, help to better understand various aspects of the tagger's performance and might be more suitable in describing the system's behavior in settings such as information extraction.