High recall, low precision, n-gram partial match

Elastic Search
1

We have setup a news search application which is returning a lot of irrelevant articles, because of n-grams and partial matches. For example, a search on MongoDB is returning articles with the word 'among' or 'ongoing' in it. These unrelated articles seem to have a score < 1. A few solutions/feature requests come to mind, but not sure which ones are possible.

We could filter by score, but there is no way to pass a filter "min_score": 0.5.

Can we increase the min_gram, max_gram settings in App Search? We like the way query_suggest works, so we wouldn't want to impact that.

We can try adding quotes around single terms like "MongoDB", but didn't find an easy way to do this in Simple-ui without overriding the existing methods/functionality.

Thoughts?

Thank you,
Carl

2

This is a challenge with App Search right now. We're currently working on solutions to solve this, but we can't share any specific dates or details at this time.

3

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