Introducing the latest updates to our powerful and ever-changing algorithms — Nebula. This Spring '22 Release brings an upgraded Vertical Ranking Model, Functions in Query Rules, additional support for multi-language experiences, and many other machine-learning improvements to optimize performance in search.
Our upgraded Vertical Ranking Model utilizes embedding technology to understand the relevance between a user's query and the verticals in the search experience so that it can display and rank verticals accordingly. This updated method means that Answers will show verticals in an even more logical order than before for each search. Plus, Administrators can influence the model for their specific experience using Thresholds & Biases. What are those, exactly?
When a user enters a query, the Vertical Ranking Model assigns each vertical a "Semantic Vertical Score" from 0 to 1 based on how relevant that vertical is to the query. Thresholds are the minimum "Semantic Vertical Score" that a vertical must meet in order to be returned at all in the search results. Biases are adjustments you can make to the final score of each vertical, which determine the rank in which they appear.
For example, a company may want to set a high threshold for the 'Jobs' vertical if they only had one job posting at the moment because they don't want this vertical to show up unless the user entered a query that is very semantically similar to that job posting.
Or, a help site may assign a +0.3 bias to 'Help Articles,' if the 'Events' vertical is frequently appearing above the 'Help Articles,' because users are primarily coming to this site looking for quick answers.