teramotodaiki I know this is a bit forward, but the mechanism that Helpfeel calls predictive intent search works really well, and in this respect, I can confidently say that it is more accurate than other approaches (though I haven’t benchmarked it). I can confidently say that it is more accurate than other approaches in this respect (although I have not benchmarked it). I really wanted to talk about it at the Tech Hour, but I couldn’t talk about it in 15 minutes.
KevinAFischer: hot take: vector embedding search is overrated
You want search over the answer space, not the question space - answer space is the mathematical dual of the question space
teramotodaiki If you’re curious, there is one explanatory slide in the second half of this slide. However, it may be difficult to find applications other than FAQ search, to be frank. It’s more of a mixed martial art that you have to manage with Human in the loop and business model design rather than an algorithm, so it won’t work just by plugging in miscellaneous information sources.
teramotodaiki: Thank you for listening to my presentation! Here are the presentation materials, please use them to find questions for the panel discussion!helpfeel_tech https://docs.google.com/presentation/d/1rj0QOumDkh4elKMPBtvYHbky6io8vMuQlVgdSrQHZHM/edit?usp=sharing…
This page is auto-translated from [/nishio/Human in the loop](https://scrapbox.io/nishio/Human in the loop) using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I’m very happy to spread my thought to non-Japanese readers.