Scoring and relevance, old article gets higher score then the new

Scoring and relevance, old article gets higher score then the new one

Marcus Lindblom

Vote:

I'm using UnifiedSearch, a simple search like this:

SearchClient.Instance
    .UnifiedSearch()
    .For(query)

I know that this example is stupid and not a real world example but it's an easy way to reproduce the error.

I have two identical pages, except that one page is older (Document.PublishDate and Document.UpdateDate) and of course the url segment is different. The older page gets a higher score which must be wrong? I tried to boost the date but it seems like that's not possible when using UnifiedSearch? Is it possible to alter the default behavior so new articles are rated with a higher score?

- Marcus

#151873

Aug 09, 2016 13:33

Per Magne Skuseth

Vote:

Check the differences on the pages in the episerver Find explore view. The date should not matter. The url segment could affect the score though.
Also, consider using UnifiedSearchFor instead, it will search in the default field, such as SearchTitle, SearchText, SearchSummary, SearchAttachment.

However, it you want newer documents to get a higher score. You could use Auto Boost

#151878

Aug 09, 2016 14:58

Marcus Lindblom

Vote:

I have checked for differences using a diff tool and it's only the dates and some guids that are different.

Ok, so you don't think that a newer date should generate a higher score? I thought so, hence my question.

#151883

Aug 09, 2016 16:19

Per Magne Skuseth

Vote:

Then the score should be the same, unless auto boosting or other boosting has been used. I'm not sure why you are seeing a different score. Have you tried specifying search fields? (InField(..))

If you really want to dig into it, you could look at the score explanation from elastic search. You could do this by browsing to http://your_service_url/your_default_Index/_search?q=SomeQuery&explain

replace your_service_url, your_default_Index with the values from web.config and replace the query with something that would make sense to you. By doing this you'll get to set the search source with score explanation.

It's in Json format, so JSONView for Chromeis quite handy.

#151884

Aug 09, 2016 16:34

Marcus Lindblom

Vote:

I have submitted an issue to the Find team with the data I extracted. I'll do a follow up when I get the answer, thanks for the help.

- Marcus

#151900

Aug 10, 2016 8:13

Try our conversational search powered by Generative AI!

Scoring and relevance, old article gets higher score then the new one