User Keyword Search should not be Stemmed, or stemming should be an Option

[Michele Morgan]

When user records are keyword indexed, the terms are stemmed.

This can be problematic with user data as it can dilute the search with irrelevant results.

An example in the Concerto data set is a patron keyword search on the term "davis" which retrieves 12 results.

Only four of these results contain "Davis" in a name field, the other eight results contain "David".

As another example, a keyword search on "Davies" retrieves the same 12 results, none of which contain the term "Davies".

This issue would be magnified in a production dataset.

It's worth noting that the Instructor search in Course reserves exclusively uses the keyword index.

[Charles P Wiggins (cpwiggins)]

We have noticed a similar problem with the OPAC search.

[Kathy Lussier (klussier)]

Hi Charles!

Stemming in catalog searching is a little different and is configurable. It's configured from the client from Server Administration -> MARC Search/Facet Class FTS Maps. There is an entry for the author class with the Text Search Config value of english_nostop. You would need to edit that entry to set "Active?" to no. I believe a reingest may be required after this change.

I hope this helps!