If the highest ranked hit has a total greater than 1, all of the document scores are normalised to be between 0 and 1, with the highest ranked document having a score of 1. When Lucene calculates the score, it finds individual scores for term hits within fields, etc. If you are sorting by field other than relevance you need to set doTrackScores(true) & doMaxScore(true). Yes, the score will always be between 0 and 1.

Returns the max score of all matches, when doing default sort by relevance. The Lucene PMC is pleased to announce the release of Apache Lucene 8.10.1.Apache Lucene is a highperformance fullfeatured text search engine library written. More detailed explanation can be found here Lucene Similarity You can turn on debug option to see how score is computed, Ex:Ģ.1740298 = fieldWeight(text:'mmdci bldleg 02' in 210), product of:ġ3.388552 = idf(text: mmdci=812 bldleg=264 02=6220) For instance if the score of a document#1 is 1.9 & that of a document#27 is 0.8, then document#1 is far superior match than document#27 when maxScore is '2.1740298'. Whats important to remember the value of score itself does not offer much meaning, but when compare the relative score of a document to that of maxScore of the query it provides value. Lucene implements a variant of the Tf-Idf scoring model. Ex Req: look for maxScore='xx.xxxx' in your response, and this will go above/below 1.0, depending on the query, results, relevance. To find out what is maximum score for a given query, you can request score field through fl parameter, it needs to be explicitly requested. The maximum value of score depends on the query being executed. These percentages still cannot be compared across searches though. Queries will, by default, use DefaultSimilarity, which weights terms based on how frequently they occur in a document. Lucene uses a combination of the Vector Space Model (VSM) of Information Retrieval and the Boolean model to determine how relevant a document is to a. This will give you the equivalent of a percentage score. normalisedValue (x - min) / (max - min) which will give you a value between 0 and 1. Then for each result with relevance x, you can calculate. If you really do need a value between 0 and 1, you can normalise each score based on the ratio of its value to the highest score from the search. If you've sorted by relevance this should be staightforward - get the relevence of the first result (max), and the last (min).

Scores across different searches are not directly comparable. The score of a particular hit is absolute, meaning that it can only be used as a comparison to the highest score from the same search. Search for jobs related to Apache and tomcat or hire on the worlds largest freelancing marketplace with 21m+ jobs. If a document contains both A and B, the score will be a combination of scores from both sub-scorers.īecause there can be many sub-scorers, the total score can be greater than 1. Each part of this query is searched independently using a sub-scorer and given score for the relevant part of the query. This query is broken into its constituent parts - A and B. When Lucene does it's scoring, it basically sums a set of scores together to give a total score.

Apache lucene scoreing