Is there a way to use non-text fields for document similarity in ElasticSearch, such as dates or integers?

Using the more_like_this feature of ElasticSearch, I understand how to use tf-idf or other metrics to find similar documents based on text fields. However what if my use case is more like this:

#   Name    Description    Price    Date
1   A B C   Ba Bi Bou      100.0    12-01-18
2   A B Z   Ba Bi Xon      250.0    01-11-11
3   X Y Z   Xa Xu Xon      100.0    12-02-18

Based on text fields only, document #2 should score higher than document #3 in terms of similarity with document #1, since document #3 would score 0. However, taking "price" and "date" into account, document #3's score should raise.

Can I do something like that with ElasticSearch? The only documentation I find deal with text fields. Adding non-text fields to the more_like_this query doesn't cause any exceptions, however scores are completely unaffected.

1 Like

As of this moment I don't think this is possible.
From the documentation:

Important : The fields on which to perform MLT must be indexed and of type text or keyword

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.