Can Zendesk Guide search find UTF8 strings and partial word matches?

Answered

6 Comments

  • Skip Moore
    Zendesk Team Member

    I know we can't do partial word matches. But UTF-8 language should work. How are you typing in the UTF-8 strings. 

    0
  • Skip Moore
    Zendesk Team Member

    I was able to search by Japanese character 

    ガイドテンプレート 

    0
  • Hans Vallden

    Humm...I wonder if the search uses an asynchronically built index? When I tested immediately after creating an article the search didn't find my UTF8 strings. Now it does. Or maybe I'm just losing my marbles. :)

    0
  • Hans Vallden

    Nevertheless, lacking partial matches and other "advanced search features" are a big downside particularly for any language based on stemming.

    0
  • Patrick Plaggenborg

    Are you looking to implement partial word searching in the near future? Or is there a way to turn that on right now?

    0
  • Gorka Cardona-Lauridsen
    Zendesk Product Manager

    Thank you for the question.

    In instant search that is performed while typing, we do use partial word matching, but to determine the results on the SERP we instead use fuzzy search (approximate string matching) which will include partial matches. In some cases where the partial match is the desired result, they may be ranked lower than approximate matches of less relevance though. That is very query and use case dependant.

    The reason for the difference between instant search results and SERP results is that instant search needs to return results much quicker than the SERP for a satisfactory user experience. The search algorithms and resulting behavior is thus different between the different search result interfaces. For example does instant search also only match on titles, where as the SERP results are matched on several other content parameters as well.

    We continuously try to finetune our search algorithm through A/B testing tweaks and features, but we don't currently have plans to swap fuzzy search for partial word matching in SERP queries, however I will add your feedback to our search relevance feedback so we can evaluate the idea and possibly prioritise to experiment with some implementation of partial word match.

    I must admit that I need to familiarize myself a bit more with if and how you can combine partial word match with approximate string matching, so I would gladly receive input from anyone on:

    1. What is the use case that makes partial word match necessary, but where approximate string matching is inadequate?
    2. If you know of any implementations of both partial word match and approximate string matching together and could describe how the dual logic is implemented it I would love to hear .

    Thanks!

    1

Please sign in to leave a comment.

Powered by Zendesk