If case of interest in constant score queries i.e. there many other advanced functions to use in boosting, first of all you can decide if additive or multiplicative: It is currently possible to give an additional boost to specific fields, explicitly: name: We’ll describe what is currently available in your search solution, and some possible improvement there.
There is a whole world of ways to customise the Solr scoring function.īooks have been written on this topic, so describing all the ways to tune relevance in Solr is out of scope here. The score will then affect the ranking of the results (higher the score, higher the rank).Įven if Apache Lucene/Solr implements what is called an “probabilistic information retrieval ranking model” the score produced for the search results are not probabilities in the strict sense of the mathematical definition, so don’t expect scores to be 0 1.įor more information on Lucene/Solr scoring: When a search result matches the query, each term occurrence is scored and then combined to build the score of the search result. Starting from Solr 6, Lucene implements by default the BM25 similarity, a scoring function that make use of the features forementioned. document frequency of the term matched in the corpus field.term frequency of the term matched in the search result field.
Modern IR systems relies on various factors to score a document when matching a user query: Search relevance and search results scoring is a complex matter. IR Ranking Functions Introduction - Default Scoring
This page describes the SOLR search engine optimisation possibilities to fine-tune the free-text search behaviour. Autocompletion Service (potential future development).IR Ranking Functions Introduction - Default Scoring.
0 kommentar(er)
