0

Has the Search changed since V18R2?

Can someone please provide specific details of how the search acts differently in 18R2?  My team provides support for Vault and has received multiple notes in the last two days from people who claim that search terms that previously gave pretty targeted results are now being tokenized differently and many more results are being returned.

For example 123-456 is being treated as 123 OR 456.  The use of double-quotes works fine, but it appears that 18R2 may have fundamentally changed the way things work.  I think tokenization was previously a thing, but perhaps the implementation of it has changed.

So... what would be most helpful would be to see the definitive list of what's different now so that we can distribute that to our user base and let them adapt to the new changes.

Thanks in advance,

Kevin

5 comments

  • Avatar
    Jonathan Chickneas Official comment

    Hi Kevin,

    There was a change made to search in 18R2 to resolve some ranking problems we were seeing with multiple term searches. The changes we made fixed the ranking issues we were having, but required that all searches be tokenized. This is why a typical doc number pattern like ABC-DEF-1234 produced a lot of results by matching to ABC or DEF or 1234.
     
    We are releasing a patch tonight that will continue to treat doc number search patterns the same way we did in 18R1, without sacrificing the ranking improvements we made for 18R2. You should start seeing single results again for these types of searches after that patch goes out so your users will not need to adapt to the new behavior.
     
    Thanks,
     
    Jon
  • Avatar
    Kevin O'Brien

    Still wondering if there is any information that can be shared about the impact that 18R2 had on the search behavior.  More of our users have noticed their typical searches act differently now and are asking questions.  I'm hoping to get some details that I can share so we can tell our users how to adapt to the new behavior.

    Thanks,

    Kevin

  • Avatar
    Kevin O'Brien

    Thanks for the information, Jon.  To make sure I'm clear, you mention "doc number search patterns" will be affected by tonight's patch.  Does that only impact the "Document Number" field, or are all alphanumeric, hyphenated values in any field impacted.  I ask only because we have the following fields that use this type of pattern:

    • Document Number - the out of the box field
    • Name - free text, but people use their own manually-generated patterns like ABC-123 or 123-456
    • Test Substance and Regulatory Application - object fields where the Name of the object as well as other object fields may have alphanumeric IDs and patterns

    I'm sure there are others, but the Name and the 2 Submissions fields are the ones that are generating the most questions in the past week.

    Kevin

  • Avatar
    Jonathan Chickneas

    All alphanumeric, hyphenated values in any field will be impacted (Name, Test Substance, Doc Number, etc). If the only term a user searches for is alphanumeric or contains punctuation inside of the term, we are treating that search as a lookup to a field value that will not be tokenized. This means that the user will only get matches on doc/object fields that contain the complete search term rather than getting partial hits. This is how it behaved in 18R1.

  • Avatar
    Kevin O'Brien

    Thanks for the clarification, Jon.  We'll give it a shot tomorrow and report back if we have additional questions.

    Kevin

Please sign in to leave a comment.