Relevance Calculation

Regarding relevance, the calculation done for relevancy score is not direct and is complex, it is based on field length norm, term frequency, and many other factors (which cannot be controlled directly). The field length norm is based on the size of the data of the field. Also, the custom field included in the search would have a separate weight. 

The relevance calculation is based on Elasticsearch. Regarding relevance explanation, please refer the links below:
https://www.elastic.co/guide/en/elasticsearch/guide/current/scoring-theory.html
https://www.elastic.co/guide/en/elasticsearch/guide/current/relevance-intro.html
https://www.compose.com/articles/how-scoring-works-in-elasticsearch/

From 8.6.x onwards we have tune parameters available in SearchBlox, using which it is possible to calibrate the relevancy score. These tune parameters are based on the following fields

  • title (tune.0)
  • description (tune.1)
  • keywords(tune.2)
  • url(tune.3)
  • date recency (tune.4)
  • url length (tune.5)
  • number of clicks (tune.6)

eg: tune=true&tune.0=5&tune.1=3&tune.2=3&tune.3=5&tune.4=180&tune.5=10&tune.6=10

When tune values are specified for each of these values as above, Elasticsearch uses them to calculate the relevancy score where the weights are multiplied to the score i.e., the factored tuning parameter value will be multiplied to the score calculated in Elasticsearch mentioned earlier. 

Note: The score will not be based on tuning values alone (even if tuning is enabled)

 

 



To learn more about SearchBlox please visit our developer document site https://developer.searchblox.com/docs

Have more questions? Submit a request

Comments