@Pierre Tapia , @Shawn and @Jared -- you all asked for improved accuracy from Storytell:
We want to let you know that we've updated how we're getting summaries so that we can deliver more accurate results. Previously, we captured information from the summary of a summary, which led to data loss. Our new approach selects the relevant Story Tiles™ and uses their raw text. What we're seeing is that this improves accuracy -- let us know if you agree!
Watch Parker, one of our engineers, describe how we did this:
for our crew: the JTBD is "improve accuracy of chat messages related to source content".
This is a phenomenal update! I had some “trust issues” with StoryTell due to several false-negative responses regarding specific terms, despite the explicit mention of those terms throughout the source document. This looks like a foundational improvement that should clear those mishaps up. Great work StoryTell team!