Controlling Score on Phrase Suggester

Hello, ES community and devs. After expending a considerable time with the phrase suggester, I've come to a point where I need to further control it, specially on the phrase scoring part of it.

Besides the smoothing models, are there any plans to allow us to further control the score of the query?

Perhaps, by allowing us to set a string_distance parameter on a per-term basis, just like in the term suggeter?

Or by having a way to "boost" the terms generated by each of the sections in the direct_generator (i.e. the terms generated by the first direct generator to have a boost factor of 2, and the ones by the reverse generator a boost factor of 1) ?

I do not know if this goes down to the lucene implementation of the suggester, or if this can be tweaked at the ES implementation level.

Please comment if you have experienced similar issues or wish this kind of cuntionality.

there is nothing in the pipeline but you are happy to propose changes and open issues for those. usually it's a good idea to explain what the problem is rather than proposing a specific solution first.

Hi, Simon. The issue I'm facing is that the suggester seems to be scoring based on document matching instead of based on the edit distance, thus forcing me to set the max_edits generator parameter to 1.

If my assumption about the suggester scoring is correct, then this would be a query that should have a few tweaking params :slight_smile: .

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