The so called learning search algorithm

Hi there,
I represent a trade company which has a product catalog of around 150,000 products. I'm a big fan of elastic search but the marketing people are fans of SLI Systems. The SLI Systems people have this promise of a 'learning search algorithm'. I suspect what they are referring to is some way that they can automatically add some search terms that people use that somehow bring them to the desired search result. Or perhaps that ranks some results higher based on popularity for example.

My questions are:

  1. How solid is this concept? Is it even possible to auto populate search words without contaminating future search results? For example, some user would think that 'shower base' means 'a base of a shower' while another user may think that 'shower base' means 'a shower from a brand called 'base' (this is a real example).
  2. Does elastic search do anything similar to this where it can be dynamic (or elastic :slight_smile:) with the data that it holds in the index?

Would love to get some feedback on this.

Thanks,
Tamir

It may also be http://opensourceconnections.com/blog/2017/02/14/elasticsearch-learning-to-rank/, which you can obviously implement with Elasticsearch :slight_smile:

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