Controlling the Results that Matter to You
by Kristina Drye, on Apr 8, 2021 8:22:25 AM
As we discussed in the first blog post of the series, one of the most important updates to GOST in the Spring 2021 GOST Release is the change to content-level reliability (CLR) and entity-level reliability (ELR) scoring.
The combined improvements of CLR and ELR put the user firmly in control of how many false positives they choose to see for different use cases. Using the slider bar illustrated below, an analyst using GOST can choose the weight of returned information to be more closely aligned to the right person, meaning that false positives would be minimized; or to be more closely aligned to the right topic, meaning that false positives might be a little higher, but you get more information about the topic you are looking for.
This is important because different analysts have different priorities for different tasks. An analyst that is intent primarily on reducing false positives and finding the correct entity will want to weight the returned information farther to the left of the scale bar
But an analyst that is interested in not only entity resolution but also assessing completely whether or not there is a risk or threat, even if some of the returned information has medium-level reliability, would want to weight the returned information farther to the right of the scale bar.
With the GOST Spring 2021 Release, both analyst needs can be accommodated, and any in between. Additionally, if it were of benefit to the analyst, it is possible to run one set of entities with one scaled weight; and then re-run that same set of entities with a different scaled weight. A comparison of the two results would also be useful to analysts with multifaceted entity-resolution needs.