Skip to main content

Will time stamping my data help serve up the most relevant data?

FeedbackQuestions and answers

Works and gives answers from years of our Slack conversations converted to text and uploaded to the chatbot.

Issue

If I ask the bot a question I want it to give me the most up to date answer.

Example

Our slack data might say 100 times

"admins in our software have 10 super powers"

I have uploaded new data that says

"admins have 15 super powers"

I can not manually go through all our data and remove old data. I am trying to figure out a solution.

Will time stamping our data help at all?

2020-01-05 10:01 - admins in our software have 10 super powers

2021-02-05 10:52 - admins in our software have 10 super powers

2022-03-05 10:52 - admins in our software have 10 super powers

2023-04-05 10:52 - admins in our software have 10 super powers

2023-12-05 09:54:47 - admins have 15 super powers

Posted by Zack Mertz ·
Status: Rejected3 comments

Log in to comment and vote

Comments3

  • My AskAI

    Team•

    Dec 21, 2023

    This is a very good point that is actually quite hard to do unfortunately, and I am not sure your solution would resolve it.

    The issue is with how the search for an answer works (using semantic search), which doesn't really place any weight on dates/times (as they have no semantic meaning).

    The other issue then is about the date of the content coming in, as what should be used as the reference date for a piece of "latest" content - the date it was added, the date it was updated, the file name date of the content, the meta data creation date of the file, the meta data update date of the file - all could be valid in different circumstances so an AI figuring it out is very difficult (as it would be for a human!)

    So the (unhelpful (sorry)) answer, is just to keep content up to date in your knowledgebase and remove dated/inaccurate content.