Search with useRelevanceScoring

GraphSearch ranking has been overhauled to prioritize results based on extracted concept relevance scores. Now, users see the most pertinent documents first, ranked by precision scores assigned during concept extraction. This approach surfaces the most valuable information immediately, even for complex, domain-specific searches, eliminating the need to sift through pages of loosely related results.

This sophisticated ranking mechanism ensures that when searching for complex or domain-specific topics, results are prioritized not just by keyword frequency, but by their deep semantic connection to your taxonomy, leading to faster and more precise insights.

In a search request we use a SearchFilter's value field to specify concepts from our taxonomy to search for and the boosting  field to indicate how much to boost a search result with that value. When multiple search filters are provided to specify multiple concept values, setting the useRelevanceScoring field to true can improve the ordering of the search results.

With this on, if the search finds a document that has a set of the values that were specified in the request (or, according to a taxonomy, related values such as plural forms and alternate labels), a score is calculated using the base score from Elasticsearch multiplied by each value's boosting factor. Then, to calculate the document's score for ordering of search results, all the scores are added together for that document for each value that was present in both the document and the request.