Episode 186 – Azure Synapse, Cognitive Search, and Forms Recognizer

In Episode 186, Ben and Scott talk through several scenarios around the use of Azure Synapse, Forms Recognizer, and Azure Cognitive Search. Transcript Email Download New Tab - Welcome to Episode 186 of the "Microsoft Cloud IT Pro Podcast," recorded live on July 9th, 2020. This is a show about Microsoft 365 and Azure, from the perspective of IT pros and end users, where we discuss a topic or recent news and how it relates to you. In this episode, Ben and Scott dive into some work Ben has been doing lately, involving Azure Synapse, Cognitive Search and Forms Recognizer. I don't know how many of our listeners follow American football. Did you see the news from the American Football League or the NFL? Technically the National Football League, but they only play in America, so I don't really know why it's national, although, international, I don't know. Anyhu, did you hear the news from them yesterday? - I did not. - NFL teams will be forbidden from post-game interactions within six feet of each other and jersey exchanges between players will be prohibited during the 2020 season. The reaction from the NFL fan or NFL players, "Aren't we going to be touching each other's jerseys "during the game?" Kind of pointless to prohibit interactions and jersey exchanges after the game. Or this is a perfect example of the NFL thinking in a nutshell. Players can go engage in a full-contact game and do it safely, however it is deemed unsafe for them to exchange jerseys after the game. - Yeah. - Let that one sink in, because you hear stories about what happens in those games. I'm sure those players are getting spit on and get spit in their face when they're under those piles. My random tidbit for the day. - Yeah, I don't know what to tell you about that one. Just a little bit of just logic that does not flow through that app. - Yeah. Okay, so there's my random tidbit for the day. We'll keep it short, we won't go on a 10-minute side topic. We've been doing that the last few weeks, only about Max stuff. - Oh, we could keep going about Max stuff. - We could keep going about Max stuff. We could turn into the "Microsoft 365 Apple Podcast." - Yeah, we could talk about how apps crash and eat up the GPU in wrong ways and just all sorts of things. - Or we could talk about running betas. - We could talk about running betas and things that go on there as well. No one should run a beta on their daily driver, but here you and I are. But you go further than I do, you do the operating systems on your laptop and PCs as well. I'm not that brave. - Yes, I have not done my operating system yet on this go around. I have learned my lesson after the last two or three. I'll probably wait another week or two. - But you did your watch. - I did my watch. - Why? - It's like, 'cause what harm can it do on a watch? - What harm can it do? Let's go back and pull up the thread. Hey Scott, isn't it annoying how I can never update the watch 'cause I have to reset it every time? What harm could it possibly do? Pushing two and a half gigs over a crappy Bluetooth LE connection, yeah. - It only takes four hours to update and then once it updates, it doesn't have enough space to install the next update. - You should not do that. Your phone, your iPad, sure, you can back up and restore those. That's your other problem with your watch, is you have to take it in for service if you wanna step back in an iOS version. - Or if you wanna go backwards. True - Yes. - I can always do a restart though after I upgrade, once I reset it, just to go to the next beta version. All right. - That is one way to live your life, for sure. - On the edge, or being very patient while my watch updates. I mean, I figured with my watch, what do I? It really has to keep time. If I get different notifications, all of that, to me my watch was less risky because there's really nothing I could lose on my watch. I mean, I suppose I could break it, but I'm not gonna incur data loss if a beta version on the watch goes bad. - Yeah, but you won't be able to track how many times you've stood up. - See, I don't stand up. Like this is me sitting and this is me standing up. My watch can't really tell, 'cause I stay the same height, no matter what. For all of you that are listening on audio and couldn't see the video. - I am trying to help you out here and you just refuse, you're obstinate. - I am. Other people have told me that, or I'm just stubborn. I'm a firstborn, I am stubborn. As IT professionals in the cloud era, sometimes it feels like we don't speak the same language as the rest of the organization. So when stakeholders from finance or other departments start asking about a specific project or teams that Azure costs, they don't always realize how much work is involved in obtaining that information, sifting through cluttered CSVs and a complex, massive metadata in order to manually create custom views and reports. It's a real headache. On top of helping you understand and reduce your organization's overall Azure spend, ShareGate Overcast lets you group resources into meaningful CostHubs and map them to real world business scenarios. This way, you can track costs in the way that makes most sense with your corporate structure, whether it's by-product, business unit, team or otherwise. It's a flexible, intuitive and business-friendly way of tracking Azure infrastructure costs. And it's only available in ShareGate's Overcast. Find out more on sharegate.com/itpro. So back to cloud news, Azure. Should I talk about Azure? I was doing an interesting project this last week 'cause I was looking for projects, and this one popped up and I figured, kind of like my Apple beta watch, why not? What could go wrong? - Yeah, what could happen here? - I'm a modern desktop guy, what could go wrong if I start doing something related to development in Azure? Under the story, it's all working, absolutely nothing. But I spent a week. And by a week, I mean like a full week, like Monday I worked on it from 8:00 a.m. until 4:30 a.m. I almost worked on this for 24 hours, and then I slept for a while. But I got to play with Azure Synapse Analytics, which, in parentheses, formerly SQL data warehouse with a question mark. Oh no, that was the what is, it's the beginning of a, apparently I should have slept more last night. Anyways, Azure Synapse Analytics, Azure Cognitive Search, some of the different features in Cognitive Search, and then I tied that in with an Azure function, with Forms Recognizer, and kind of built out a demo lab-type scenario, leveraging all of those different tools. And actually, it was really interesting. I think part of the interesting thing about this one for me, is it's looking at all these Azure services, which under the covers, you're like, "Oh, that's modern desktop stuff. "They just rebranded it and made it user-friendly." - Sure, that's one way to think about it. - At least for some of it. So should I walk through the scenario that I built and then we can kind of talk about some of the tools? - Yeah. - Okay, so this is how it worked. I took some Azure Blob storage. So first of all, we have a storage account, Azure Blob storage. We've talked a lot about that. So I didn't include that in the whole line of processing, but in this Azure Blob storage, I created just Blob storage, upload some invoices to it, so some PDF invoices. And Cognitive Search has the ability to go index this Blob storage, so I set up the Cognitive Search, goes and indexes this. And you can do some different things with Cognitive Search in terms of how you extend that indexing process to pick up different data. So the way I did it is when the index ran, it incorporated a custom skill set which actually went off and called an Azure function. So upload invoices, run an index, call an Azure function with this custom skillset and this Azure function passed over the necessary information for Cognitive, or, not for Cognitive Search, for Forms Recognizer to be able to pick up these PDFs as they were getting indexed. So Forms Recognizer picked up the PDFs, did a bunch of processing and these PDFs doing Form Recognizer to pull out things like prices and invoice numbers and quantity and the total amount of data, so all that information you may wanna pull off an invoice. And then another part of the whole kind of search is knowledge store. So you can take these pieces of metadata and actually shove them out to enrich your search results or, in my case, I stored them as JSON files back in my Blob storage that you just go in and pick and choose key pieces of information. So I was saving like five different pieces of information pulled off of these forms via that custom skillset to JSON files in Blob storage. I think I ended up setting it just to once a day. My pipeline ran in Azure Synapse, that actually went in, picked up these JSON files via a data flow in my Synapse pipeline, pulled out the key pieces of metadata, did a little bit more data manipulation, did some string cleanup, did some converting from strings to integers or to decimals, and then pushed them into a SQL database or the Azure SQL pool that was in Azure Synapse. So end of the day, it didn't do a lot of the data warehouse aspects of Synapse or some of the analytics. This was kind of part of a bigger solution as well, that incorporated a bunch more into Synapse. And there's a bunch of things you could go do with that in Synapse from there, but that was kind of the gist of how all of those tools ended up tying together in the solution. - Yeah, I think it kind of speaks to that ability to link services together and kind of build out the glue wherever you did it. And then you have a bunch of extensibility in there as well, right? Just to cross that stack to be able to, the thing you said you're not using potentially Synapse as a true d