Boost
The purpose of this widget is to give you an easy way to boost specific documents in the search results. For instance, if you are a retailer, you might want to boost products with a good margin. However, your boost options are restricted to the data available.
Pipeline
At the top of the widget you specify the search handler that will be used in the widget.
The changes you make in this widget will be applied to the selected pipeline.
The pipeline called tellusrSearch implements the /tellusr/api/v1/query operation.
Boosting rules will be applied as the final score transformation to the doc list, that is, after merging semantic and regular scores. Boosting rules only re-rank documents on the first few pages of the hit list. Boosting rules are ordered; first field boosting is applied in the order that they appear in the GUI, and then age decay is applied in the order that the rules appear in the GUI.
When checking out queries in the dashboard, the total boost effect of all rules is displayed as:
The boost value that is reported is the difference of the new score (after applying all boost rules) and the old score.
Boost Fields
Underneath is an example from a book store where we have specified that we want to boost all books by the author “Stephen King” by adding “0.5” (additive boost) to the search score of the documents that contains the text “Stephen King” in the authors-field. The score prior to the boost will be modified as:
new score = old score * multiplicative factor + additive term
When you add a new boost rule, remember to click the save icon at the right of the rule in order for the changes to take effect. If you want to delete a boost rule, you can do that by clicking the “x” next to the rule and then save. Changes only take effect when save is pressed, and not during editing.
If you want, you can specify that the rule should only apply for a given period. You do that by selecting a start and end date for the rule.
Prioritize Recent Documents
If your documents are dated by a field such as date, you can set up boosting that prioritizes new articles. The search score of an article is modified by a boost that is subject to half-life. At document age = 0, the boost is maximal, and at document age = half-life, the boost is half of that. The magnitude can be adjusted to modify the effect of age decay. Documents are re-scored by the following operation:
new score = old score + magnitude * decay
where decay is determined by the settings and age of the document.