Score parameter and prioritization in Data Fusion API
Colleagues,
We are working with entity group requests (such as entity/analyze/connect, entity/search etc) and want to understand how 'score' parameter works. What are the min and max values here and how this score is calculated.
The other part of the question is how API decides which entities to return when we limit it's number in request. Do we receive connection in random order (which means most relevant connection can be returned after several requests)?
The entity score is generated by Solr, see https://wiki.apache.org/solr/SolrRelevancyFAQ for more info. In general, this is best used when doing an unstructured text search.
Score on edge annotations is user supplied, and in this case derived from TRIT relevance. See TRIT's user guide at http://www.opencalais.com/opencalais-api/.
The entities are returned in no predefined order, i.e., random. Correct, you would need to see all connections to verify which one is the most relevant. At the moment the call is optimized to find a connection but does not optimize for relevance.
We're working on adding an order to the resultset based on timestamp, i.e., newer paths would be returned first, but this feature is not yet in the product.