Relevance
Relevance |
|
|
In this section, it is possible to increase or decrease the relevance given to each of the criteria that will be taken into consideration in the full text search.
Default relevance table
To adjust the value of the relevance criteria, use the slider control to define the desired value or fill out the field available next to the control. It is possible to set a value from 0 to 50.
How does the relevance work? The result sorting by Elasticsearch is obtained through a similarity score via Okapi BM25 algorithm. In this algorithm, the following factors are taken into account to calculate the score: 1.Frequency of the searched token in each of the indexed fields. 2.Size of each indexed field with the token. 3.Number of documents with the token.
A mathematical calculation is performed uniting these three factors. For example, a document with the "bucket" token appearing twice with a size of 1000 words is more relevant than a document with the "bucket" token appearing three times with 10000 words.
That is because, in a text with a lower number of characters, the size of the document may be more relevant than the frequency (obviously, if the token frequency increases, the analysis will be reconsidered, as it is a sum of the three factors).
However, if a text has a high frequency of a word (especially for smaller texts), the relevance of this token will be smaller for this document, as the algorithm considers it a "common token" for that document.
Moreover, after obtaining the algorithm score, it is multiplied by the value of the field previously defined in the relevances.
Thus, it is important to check if there are attributes with the token in the searched document, since they will also increase the search score.
|