Search Score ratings are too low

Description

When a user conducts a search in the blogs portlet, the search score ratings are very low, even if search keywords match exactly with a blog entry's text fields such as the title, content, and/or tags.

Here are the steps to reproduce the issue:

1. Add blogs portlet
2. Publish a blog entry with "test" as the title, "test" written in the content, and "test" added as a tag
3. Type "test" into the search field
4. Check the score rating. The score rating will likely be valued as 0.3 out of 5.

Environment

Tomcat 6.0.33 + MySQL 5. 6.0.x Revision: 94267. Tomcat 7.0.22 + MySQL 5. 6.1.x Revision: 94251.

Attachments

1

Activity

Show:

Christine PinceDecember 7, 2011 at 11:50 AM

PASSED Manual Testing following the steps in the description.

Fixed in:
Tomcat 7.0.23 + MySQL 5. 6.1.x Revision: 94938.

Jonathan PotterNovember 28, 2011 at 2:16 PM

In my opinion the solution to this is to remove the "Score" column from the search results entirely. It is more confusing to the user than it is helpful to have the best result matching 0.3 out of 5 stars even when it seems like it's a close a match as it can possibly be. There is not an error in calculation here, just a misuse of the score property. The score comes from Apache Lucene which implements the searching. The score is not an absolute and is not even technically bounded to any range. It is meant solely to be a way to order results from a search relative to each other. See http://wiki.apache.org/lucene-java/ScoresAsPercentages.

This means that if you search and one result has a score of .7 and another has a score of .3, you simply know that the first result is more relavant than the second. If you then add a blog entry and search again with the resulting scores of .01 and .005 it doesn't mean that the .7 result from the first search is more relavant than the .01 result from the second search because the scores are not absolute, they are relative. You still only know that the first result of the second search is more relavant than the second result of the second search.

Showing the Lucene score as stars or a percentage of relevancy implies to the user that it is an absolute indication of relevancy, when it is in fact entirely the opposite.

ginson.renNovember 27, 2011 at 10:29 PM

I was able to reproduce on trunk rev-94251 and 6.0.x rev-94267. I will update this ticket to reflect this issue.

Christine PinceNovember 18, 2011 at 5:37 PM

Able to Reproduce on:
Tomcat 7.0.22 + MySQL 5. 6.1.x Revision: 93832.

Fixed
Pinned fields
Click on the next to a field label to start pinning.

Details

Assignee

Reporter

Epic/Theme

Fix Priority

3

Git Pull Request

Bug Type

Regression Bug

Components

Affects versions

Priority

Zendesk Support

Created October 24, 2011 at 9:32 AM
Updated June 24, 2023 at 3:43 PM
Resolved December 7, 2011 at 10:27 AM