[Music] Gerard really great to have you today I'm excited about this subject because we're going to get a little bit deep and nerdy semantic search what is it my friend yeah great question thanks for having me Chad yeah uh semantic search yes uh semantic search is a way of getting information um to your end users and uh what we often hear in the field where are you where are you using semantic search are you using semantic search the answer is we are using it uh we're using it in uh now assist and uh it's a a way to allow for fuzzier watching and getting information uh to end users this this might be a great parlay to the second question so how does it work and maybe why does it matter yeah here it is um how it works this is where we get nerdy this is where we start talking numbers so so basically the way that it works is we have something called a semantic encoder uh these semantic encoders basically take words uh in this case in English and translate them into numbers um people often call the things that store semantic Vector databases because you're basically building out a vector for your content or your search query we're basically doing matching on the the question that's being asked and the content that we have and we're trying to figure out how similar they are uh why it matters is because we have a lot of customers that ask uh question questions that are very similar in meaning um to specific types of information but do not exactly match from a keyword perspective so when you think of can I get access to uh Outlook or can I request access to Outlook like you need to understand that get and request are very similar in meaning and provide that as a response um there are some gaps okay that but we can talk about that a little bit later all right that all right so I appreciate that so with that in your experience because you You' been doing this a lot with customers in in the field what does success look like for our customers yeah in in this case this this idea of having uh semantically similar content success looks like uh our end customers not needing to perform a lot of heavy lift on editing their knowledge based articles or their catalog items and really having uh a more natural experience with the content without needing to do a lot of tuning uh and tweaking so uh what we've seen in the field is that when we've offered these semantic capabilities as part of now assist there is an uplift of roughly 5% on our recall scale which doesn't sound like a lot I know but it actually is when you consider you know the thousands upon thousands upon thousands of requests that customers get per day 5% to your point 5% doesn't sound like a lot but this ends up increasing exponentially so I got a final question for you you mention gaps so what is on the road map for this yeah the the biggest the biggest issue you can say with semantic technology is that uh we're using an encoder like I mentioned earlier and there are some words that the encoder just doesn't know um so it just makes up a number um for that for that term so you may have and it does the same thing for acronyms too so you may have business specific terms or acronyms that you're using um and we're basically going to match or have strange matches on those um so in the road map we're combining our sort of Legacy keyword approach which includes a lot of the synonyms and business specific terms or even company specific terms that can be configured by customer customers and combining that with semantic technology so what end users will end up getting is the best of both worlds and a unified sort of representation of what that looks like that is awesome we only have four minutes and four questions I know you want to get a little bit deeper I we have an opportunity if there's any questions concerns want to get a little bit deeper please reach out to Gerard myself in community you will see the URL at the end of this and with that Gerard I want to thank you for spending your time with me this was awesome all right thanks Chad