The .NET on AWS Show, featuring Diana Pham

🎧👉 https://www.art19.com/shows/net-on-aws/episodes/40c9ed2d-bfb8-42d9-8e79-cd136a8c4e2a

Brandon Minnick 1:07

Hello, everybody, and welcome back to another episode of the dotnet on AWS Show. I'm your host, Brandon Minnick. And normally with me, we would have my amazing co host Francoise. Fortunately, Francois came down with what we think is COVID. So he's resting up, Francoise. We hope you feel better. We love you, man. Take some time to rest up. But that's okay. Because we have such an amazing guest today. We're gonna bring her on right now. You know her as a developer advocate and app right hero. She runs a meetup group called Denver Food and Tech, and is competing in the upcoming Miss Colorado for America Pageant, leveraging that to empower women in tech, and is even creating your own false false lash company called AIDA love lashes. Dinah fam. Welcome to the show.

Diana Pham 1:59

Thanks for having me. That's an autobiography. And I appreciate that you were able to like go through all of that. Because I was just like, oh, yeah, yeah.

Brandon Minnick 2:13

Thanks for coming. And thanks for joining us. For folks who maybe haven't met you yet. Outside of that. Who are you? What do you do? Yeah,

Diana Pham 2:21

I mean, I think that was a pretty good summary of Diana Pham. But I live in Denver, Colorado, I have been moving around a ton. I started developer advocacy during grad school. So I think that was like three or four years ago now. Holy crap. Yeah, time just kind of flew by. But this is what I do. And for those, those of you who don't know, Brad and I, we were both at Casey, DC last last month that we call it crap. And that's where we met like a year ago.

Brandon Minnick 2:53

Yeah, that's right. And we're chatting before the show. And Dan asked me, she's like, how do you how do you find guests for the show? I was like, oscillates, you should I just meet awesome people at conferences. And I'm like, that'd be great on the show. So hopefully, they say yes, and you did. Thank you. Thanks for joining us.

Diana Pham 3:11

Really no reason why I couldn't or shouldn't have. So yeah.

Brandon Minnick 3:14

Aside from some home technical difficulties, Dan is dealing with Yes.

Diana Pham 3:25

The entrance to my home, and I'm literally sitting right in front of it. Because all my boxes are right in front of me. I had to emergency evacuate my whole situation. And that was not great. But yeah,

Brandon Minnick 3:39

right. And that's something you ever want to hear. Especially when you you're not really in control of it. Right? Like, it's, it's somebody else's fault. It's somebody else's problem. But you have to deal with the consequences. And that's, that's the worst.

Diana Pham 3:53

Oh, for sure. And workers like last year, he was like, there's one common denominator amongst all of the moving situations that I have. And it's me. So when I was a newborn. And the first thing it was the worst thing, and I don't know if any of you have had bedbugs, or know how they travel, it's pretty much really common if you go stay in hotels, or take public transit, and these thugs are gonna latch on to you and come home with you. And so I had just got back from that conference and code mash. So I was traveling back to back those two weeks, and then I came back and I had them. So then I felt so bad thinking that like I had brought him But thankfully, that wasn't the case. I found out this woman like next to me. She had them and it just spread to the entire complex. Yeah, I mean, it made me feel bad because I was I infected New York.

Brandon Minnick 4:58

And those are terrible. They're so hard to get right Have they're terrible for you? Yeah, that's yeah, I would absolutely move. If I was in an apartment complex. I got bedbugs. It's

Diana Pham 5:11

just move within the state. I had to move. Like, yeah, not even next door like I moved halfway across the country just to get away from them. But it worked out. I'm here. Denver's been great to me. I picked up snowboarding last season and I am not doing any flips, tricks, jumps anything remotely close to that, but it's it's been great being around the mountains. Oh,

Brandon Minnick 5:36

cool. It's so much fun. Yeah. I love skiing. You know, I live in the Bay area here in California. So I've got Tahoe, which is about three hours away. But yeah, I still love to go out to the mountains in Colorado and goes go skiing. I was up in Winter Park a couple couple of weeks ago. It's summertime. In the winter. Oh yeah, cuz you guys have that cool Amtrak train. Now that'll take you straight from Denver to Winter Park. No stops. So you just as long as you catch the train, there are some people that were like running after it. Like Wait, recovering. Like, no, it's a train. You can't stop it. It's already left. trains leaving the station. Yes. But yeah, as long as you can catch the train, I think it leaves like 7am. And yeah, you get a full day skiing in Winter Park. And it's incredible. You'd have to drive you have to watch traffic. All the all the pain points. God.

Diana Pham 6:30

I actually didn't know about that. But at the same time, I haven't been to I just don't know a whole lot about Denver. And why it's because I'm a developer Seth Alper advocate for places that we are in my case rent I pay rent for places that I don't live storage unit.

Brandon Minnick 6:48

Yeah, right. I I knew a guy that No, single guy. So he's like, why am I paying for an apartment that I'm never at? like nobody's ever using it? And would just use the I think he was big into Marriott points. So you're just uses Marriott points to stay in a hotel anytime he was home, and then just kept traveling. So it's not not a great life. But he did it

Diana Pham 7:16

are really good at the Nomad thing. I'm just curious, wherever everyone is from if you want to drop it in the comments. I'm just very curious. Like, to any of you have this, like character development type?

Brandon Minnick 7:32

Has anybody been chased out of an apartment for? Let us know. But I know you. You've gotten some, some samples of demos and things you wanted to chat about today and show off? Tell the folks what, uh, what are we chatting about today? Yeah.

Diana Pham 7:47

So once upon a time, like two years ago, I had been no, it's been longer than that. I don't know. I don't know. timeframe. Because of COVID. Everything's just been a blur. Of have, I'm passing by or when things are coming up because of it. But a couple of years ago, I started coding in C Sharp because I was interning at a company where I was localizing documents. So that was the very first time I got exposed to language. And I was like, oh, it's pretty close to Java, because that was one of the first languages I learned. And I kind of ran with that. And here and there, whenever there is a need for a C sharp, written dotnet themed blog or just top type of demos. And I'm super, super happy to build that. And so when you reached out to me, I think that was like a week and a half ago, I was like, Yeah, let's go because at the moment, I had a project that was built with another cloud provider. But in fact, I actually ran that during KCDC, the year before, if you weren't there. That year, I ended up doing a talk called how to talk to women. It was a soft skills talk, where I just highlighted, you know, all the harassment, and just difficulties a lot of women in tech have been dealing with just since ever, even though I was early in career, but since then, I decided to do a talk and just like, you know, promoting awesome women who have done awesome things, but just don't get credit for it, or they've gone through things that like I have not I could share with community. So at the very end of that talk, I put a Vonage phone number on the board and are like on the projector. And I let people text it. I was like text, whatever you want. And what it does is it returns Yeah, first like, yeah. Can you see my logs? And I was just like, No, go for it. Do whatever you want. So I have this number on the screen for people to text and in return, they will be receiving a sentiment analysis report of how harassing it sounded, how offensive it was. And I think there was like one other criteria. And that was super cool. And so let's Like I started digging into AWS products that currently do the same thing. And then I got this mold evacuation problem. And so I was off my computer for some time, but I was like, No, it's still pretty. I'm still going to do this, I still want to learn about AWS. And so I went ahead and I found AWS comprehend. And I don't know if you saw if any of you are following me on Twitter, my handle is Diana's oyster. But I posted yesterday about how I was at a model casting. Like full disclosure, I am not a model. But I got to be really cool to just try it out. So while I was waiting for my turn, to walk, I pulled up my phone and started reading, comprehend documentation.

Brandon Minnick 10:46

And so I think that probably everybody in the waiting room was as well. Yeah, exactly, exactly.

Diana Pham 10:55

But they're listening to music. So they can, you know, play it in their head. But no, they're all listening to AWS podcasts. They're actually listening to the show. while they're waiting.

Brandon Minnick 11:03

They all subscribe to the dotnet native a show.

Diana Pham 11:06

That was why they have their phones out, you know? Yeah. So while I was there, I was just like, well, let's do this. I'm, I still really enjoy the fact that I'm learning about this. And I just wanted to share it with everyone. And so I asked Brandon, I was like, Hey, have you covered this topic? I probably should have asked you before. But I asked him, I was like, Hey, I've covered this topic often on your podcast. And I guess the answer was no. Or you're just trying to make me feel better. But that's okay, too. So, here we are, I just wanted to Yeah, I just kind of wanted to share my findings with everyone of everything that I found out about AWS, what kind of interested me and all of this. And hopefully all of you find it interesting, too. If there's something like fun or exciting that you've learned, or yeah, you learned about AWS comprehend, while you were building your own projects, like Sorry, projects, please let me know in the comments, I would love to try out those features that I might have missed. But so just some context and Brian and you feel free to correct me wherever I'm wrong. But well, I read is that AWS comprehend, it's a service that offers like NLP capabilities to make sense of unstructured textual data. And so it uses machine learning to get insights and relationships in a text. So from things like, you know, in my case, text messages, or social media posts, or documents or like documentation, was that a pretty good? Was that a decent explanation of what it is? Yeah,

Brandon Minnick 12:37

um, you know, it's, it's so cool that we're talking about this today, because I do have a couple of conference talks queued up, because I call it I don't know if anybody else has called it this, but I call it machine learning as a service. Because essentially, it's these pre made models by people who know a IML way better than me, which is, I mean, it's not hard if you know, a little bit, you know, more than me. But you know, I'm a developer, right? So I make mobile apps, I publish them to the App Store. And awesome that all this AI stuff exists. But I it's not what I do. Like I make mobile apps. And so services like Amazon comprehend, super smart people have already built these models that can do text sentiment, or detect faces and photos and and they give us these models, essentially, for free. Where Yeah, we can just in this case, like upload our text, and it'll give us back all the info that it cannot it. And for that it's just so great. Because before these existed, companies had to make their own, like, Facebook would hire an army of developers to do their sentiment analysis for posts and help with their, you know, reporting and kind of security aspect of the platform. And, and here we are like, yeah, you threw this together in a weekend. Which is, if you go back 10 years ago, and tell somebody like, oh, yeah, I made this app that detects the natural language sentiment over the weekend, like, what are you some kind of wizard genius? But yeah, now like I said, I call this like machine machine learning as a service where I all you have to do is make a post request to an API, and I get all this machine learning benefits. So for developers like me that just want to write apps write code. I love it.

Diana Pham 14:40

Yeah, honestly, when I found that this was a thing, of course, I have my not like, not my doubts, but I just really didn't understand how simple it was until actually with documentation, I tried it out myself. If any of you want to try it out yourselves. You can also find it on the AWS website. I was a big fan and I was I think something that really drew me to it. Of course, when I had such a short amount of time to come up with something, once again giving my own my borderline homeless, you could do it all on the console. And so that's something that I wanted to highlight today and kind of walk people through if they haven't seen it already. These features are super exciting to me. And so I want to share them with you.

Brandon Minnick 15:20

If that's cool. Love it, shall we? Shall we dive in?

Diana Pham 15:23

Yes, let's do it. So let me share one of my screen real quick.

Brandon Minnick 15:30

And yeah, well, Dan is bringing up her screen. I'll also mention we have AWS recognition, it's spelt like recognition. But instead of a C, it uses a que so I'm sure if you Google AWS recognition spelled, normally, it would still pop up. But recognition is another similar service, but more for like, recognizing things in photos. So today, we're talking about comprehend, which is, as you can see on the screen for natural language processing, Text Analytics, if if you want to go a step further with photos and recognize faces, there's also a dubious recognition. But I know headed back to you. Yeah,

Diana Pham 16:08

so sorry, you mentioned recognition. I wish I found out about that before we went to KCDC. Because I ended up doing a live live object detection workshop in Python. And so if I had known about that, before, I totally would have used that I'm definitely gonna play around with that, just because I've been trying to get into more machine learning things and just finding the services or finding the platforms that actually offer them. It's just super cool. And so if any of you end up doing cool machine learning things, or have any particular tools that you're fancying, please let me know as well. Like I just I just love finding out about these things. So here we go. We're here on the Amazon comprehend page. And like Brendon said, it's it's kind of like Mystery Machine learning as a service, and in this case, natural language processing. So let's go ahead and just launch comprehend for us, and then now that we're in real time analysis, let's scroll down to our input text. Right. So this is the one that came with us. Hello. Name of customer, your name, and then their customer or their company name credit card. Ooh, let's try this out. I'm just gonna assume it doesn't work.

Brandon Minnick 17:29

Yeah, that'd be a big Oopsies.

Diana Pham 17:31

Yeah. My intrusive, intrusive thoughts. And no,

Brandon Minnick 17:36

it does look fake it the credit card numbers like 11111 Dash 00000 Dash one comma one. So it doesn't look real. The addresses from St. Anywhere. So I also don't think that's true.

Diana Pham 17:51

Yeah. And that's a very little big, very small account on your bank. But okay. Oh, nevermind, minimum payment. I was like, Oh, I would like to think I have a bit more money than that. But I'm not too far away from this. So card number. Yes. Yeah, me. dollars, for sure. For sure.

Brandon Minnick 18:13

Yeah, right. Not sure if it's absolutely. Yeah. So basically, what we're looking at here is like a sample input that somebody might send to a customer service chat on the webpage? Because I guess they're, they're having problems with their credit card and their banking. So they're trying to get some help? And what do we what are we doing with the text?

Diana Pham 18:37

Yeah, so let's look at this particular, let's look at this particular sample. So once that's all done, you need to have anything lower than 5000 characters. I'm not sure how you work around with that limitation. But I don't know if you do on top of your head, but I'm pretty sure that there must be some type of way to work around that. But we're gonna hit analyze here. And right before analyze, or below analyze, you're gonna see the results. So yeah, you're gonna see the results. And directly under that, sorry, we're gonna look at insights. And so here you see like entities, key phrases, language, personal information, sentiment, I'll go a little bit more into what each of them mean. So and entities here, it's pretty much what is happening is that it's comprehending. Or sorry. It's like recognizing specific entities. So people places the amount in your minimum payment or the date here. So it kind of separates that out. Whatever. I'm sorry. Go ahead. Yes,

Brandon Minnick 19:47

it looks like it's underlined, all the important things like I was just reading it, and if I skip every word that's not underlined, then I see their name. I see the company that worked for the credit card number, the amount the date. And so, yeah, I don't know how it did it because I'm no machine learning expert, but somebody smarter than me figured out how to identify pretty much all the important words in this in this textbox.

Diana Pham 20:14

Yeah, so I guess like here we have this blurb. But another good use case of this is, let's say, for example, when you're filling out an application where you can upload your resume, I don't I don't know if I've ever successfully uploaded a resume that was perfect at depicting the autofill. Based on whatever is in there, you know, like, you'll have something like work experience, and then they'll put your school in there. But pretty much what this does is like, this would be a great use case where you upload your document, and they need your phone number, email, whatever it is, and then it just kind of copies over. Yeah, that's just one use case I can think of, but anything

Brandon Minnick 20:52

to improve those online recruiting, upload your resume here. And then you get to the next page. It's like, retype your resume here. Like, why? Why don't I just upload the resume if I need to write everything back? And again,

Diana Pham 21:06

I'm really curious, like, if there are any recruiters out there who can tell us, what does that look like on your end? Because, you know, we're out here spending time making our resume beautiful, or at least like aesthetic enough that someone wants to look at it, only for no one to look at it, because they're asking us to fill all those things. And then part of me is like, food, like every single Yeah, it just gets a bit much. But um, but yeah, here we are comprehend really great at that. So going over, so I go ahead, run.

Brandon Minnick 21:42

Oh, I just said nice. Yeah, yeah.

Diana Pham 21:45

So going over to key phrases over here. Here we see it just like identifying the most important phrases. And in this case, there are a lot of phrases, which I would expect, because this is yeah, like money involved. There's always a lot going on. Except that's my case. Yeah, there's not a lot of money doing much in my bank account. So yeah, here we have, you know, all these things, being under lighted once again. And a really good use case for this is like, say you want to grab some documentation, I should have done this with the AWS example. But so you've grabbed some documentation, and you're like, Wow, that's a very big blurb, or a blog. So you just copy paste it, throw it in here, and just and it'll probably help you or it could help you generate, you know, a summary. And I think that's really helpful, just so that you don't have to read the entire thing. And admittedly, I do do this more often than sometimes I'm just like, Okay, well, clearly, I'm on documentation page that's supposed to answer the question that I am having. But I have been scrolling more times when I would like, like, when this little thing reaches like four times, that's when I'm like, Okay, it's time to select all and copy and paste. So this is one really good use case there. And hop it over to language a bit more self explanatory. Here, you can tell that language it says that it's a 99%. Sure, or confidence level that is I can confirm this is English. Let me actually see what happens down here at key phrases. Oh, so actually listed out for you. And it gives you the competence score as well. And for the most part, most of them are near 100%. Here, this person see they are they're pretty confident that this is not

Brandon Minnick 23:40

credit card number 87% confidence that it's a real credit card number. Yeah.

Diana Pham 23:49

Well, there's that. That's pretty cool. Like it goes through almost everything, which is pretty nice. Pretty nice.

Brandon Minnick 23:55

Yeah, and I think this is something super important when when we use these services, because, again, like this is this is machine learning, it's never going to be perfect. But what it can tell you is how confident it is in itself. And then it's kind of up to us, the developers to choose what we want to and what we don't want to accept as a certain threshold. So I know in an answer, I've used this before. I'll do things like if it's below 85% competent, then I just say it's undetermined. Yeah, I don't let the end user know that. Yeah. Oh, yeah. This the person in this photo is happy because they're the model, the model is only 60% confidence. So say we couldn't determine maybe upload another photo. And these confidence ratings are super important for us as the developers to figure out how we want to use those, you know, how important is it that we get this right? Is it cool if we Get it right half the time two thirds of the time? Or is it something like banking? Like we're looking at here? It's like this, this needs to be right.

Diana Pham 25:08

Right. So during my machine learning workshop at KCDC, I actually explained this to people how your results are only as good as your data. So if you don't have a good data set, and you're going to end up with those, like very, either low confidence answers, or very high confidence, but wrong answers. So I actually had a hair tie, had a hair tie on the wrist, and they like, held it up to their screen after the demo, or like after their workshop, project worked. And they told me that the model was more confident that their hair tie was a been banana than the model was that they were a person. So once again, AI is telling you or machine learning is putting in front of you.

Brandon Minnick 25:50

I love that example. Yeah,

Diana Pham 25:53

yeah. And then once again, like don't trust AI in every way, shape, or form, which actually beauty just came out, I created a chatbot application, because who hasn't after open AI made that a thing? So I made a chatbot application, but I fine tuned it, I don't actually I wouldn't say fine tune it. But then I fixed up the code a bit to have it treat me or the user, as a patient for, you know, a mental health therapy session. And I went in and you know, went on my rant, and at some point, it asked me like, Oh, are you satisfied with my answer? And I said, No. And then what would you have preferred? I say instead, and the scenario is that my hamster died. So I'm sad, right? So trying to console me, blah, blah, blah. And I was like, No, I'm not I'm not satisfied with this answer of you, like trying to make me feel better. And they were like, okay, so what would you prefer me say? And I was like, I want you to tell me to go die. And what did it do? It told me to go die over my dead hamster. And I was like, another reason? Maybe you don't trust everything it says or tells you to do?

Brandon Minnick 27:07

But yeah, and seems today, you know, it feels like machine learning is getting smarter and smarter. And it certainly is, Gen AI is super powerful. But at the end of the day, it's it's just a computer. And I think it's something to keep in mind. You know, we want to personify a lot of companies are giving names to their a Gen AI chat bots. So feels like a person but it's not. It's, it's just a bunch of ones and zeros putting together more ones and zeros. And my, my favorite analogy is my friend, Seth Juarez. He's a AI ml guy. And I've seen him do some really great keynotes on it. And he describes it like, let's say I have a rock. And then I draw a smiley face on the rock. And you might go, oh, let's rock so cute. I'm gonna name it, Sam. But at the end of the day, it's still just a rock doesn't know anything about person. People. It doesn't have thoughts, it doesn't have emotions. So important stuff to keep in mind. As you know, we've we've shifted into this new world where there's AI everywhere. Yeah,

Diana Pham 28:17

and if you could find that link, if it's, if you can, briefly, I'd love to see that. Or we can share it in your tweet thread after the show. I actually haven't seen any keynotes, surprisingly, about machine learning or AI. Focused stuff. Actually, no, that's a lie. Who isn't incorporating AI or machine? Even if you can't like a very brief level, but that you know, sounds really cool. But heading over here, so we went over language, no personal information. You could either Yeah, you actually I don't know what these do. So offsets identify the location of the personal information in your text documents. So yeah, here we have all their information. What happens when it's labels? Oh, okay. So what actually labels what is in that paragraph, which is actually really cool. So I guess the offsets is the specific examples right here, you know, giving the actual entity the type. How confident it is that this here is name that John has a name? This is a credit debit number. Is it credit, or is it debit? But yeah, and then back to labels here. I can confidently tell you that they're those two are names that that one email address is an email address and all these other things are here. Typical. Okay, cool. Cool. And I'm trying to think like what would be a good use case in this situation? Oh, just like filtering out your social security number. Let's let's do this right now. So hi. And, and my social security number is 987 dash four four dash 5278. Just putting Social Security out for the world's not a real you can hope you can help, honestly. But let's analyze real quickly now yours. So clearly. Clearly here you see my name or it's confident that my name is Diana and the social security number. Yeah, okay, cool. Now when you all set it, let's see what happens. So if you were to offset this and decide that you don't want to show this personal information, then I'm guessing what's going to happen is that it'll have whatever printed thing here but not include my actual name and not actually include my social security number. Yeah,

Brandon Minnick 30:55

cuz there's that pie tab. I saw up top. Is that where it? Oh, that's where we're at right now.

Diana Pham 31:03

Yep. So that's pretty cool. So what I'm guessing is if we were actually to implement this into our own application, it would actually blur that out in whatever scenario that we would want it to borrow. I guess a good example is when you type your password in it has either the dots or the asterisks instead of your actual password. Because it knows that that's personal information.

Brandon Minnick 31:23

Yeah, certainly. And you know, if I could see this for all sorts, all sorts of purposes, where you just have some sort of filter or verification or review in place before, maybe we publish a blog post before you send out a tweet before we put anything on the internet, really, that says, hey, it looks like you might have accidentally included some PII. Diana, do you want to keep this in there? Here's, here's what I think looks like PII like this clearly looks like a social security number. Yeah,

Diana Pham 31:56

sure. So before the show, or when I booked my little slot. I was I was blessed with the opportunity to find an opening on Brandon schedule, and book pass these questions for me and one of them is like, you know, introduce yourself what you do blah, blah. My fun fact should have been like my social security number is

Brandon Minnick 32:19

not only handles PII.

Diana Pham 32:24

Yes, yes. And then over here, we have sentiment. So if you're familiar with it, sentiment analysis, and if you're not, that's totally cool. So what sentiment analysis, sentiment analysis is, is it's a technique and national neck, I cannot speak today, my goodness, I'm sorry, natural language processing, where it determines the emotional tone or the attitude of whatever text This is. So the best way I can describe it is like, it tells you whether something is positive, negative, or neutral, or mixed. And you just kind of feed it like whatever words or phrases the text is. And in this case, like, you know, Hi, my name is Diana. Blah, blah. It's pretty confident that what I said was neutral. I'm not happy about it. I'm not sad about it. It's it's not

Brandon Minnick 33:19

theirs, there's nothing positive or negative in the sentence that says my name is and my social security number is it's just just some boring facts. No motions.

Diana Pham 33:30

Yeah, and sorry, for those of you who might not be familiar with like the competence percentages, the higher the percentage is it's like the more pop Oh, actually no brainer. Explain this. I'm sorry. Yeah. are good. Everyone's caught up.

Brandon Minnick 33:43

Yep. What is the highest series the lowest? Yeah, their percentages? Actually, I don't think there ever is a one I think point nine nine is that

Diana Pham 33:54

I mean, is that Oh, really? No, no, sorry. I for some reason, thought, Oh, yeah. It wasn't even sure that this was my name.

Brandon Minnick 34:03

Oh, point nine, nine plus, I wonder how many nines it takes to add on that plus side? Yeah,

Diana Pham 34:09

yeah. Cool. And I've only targeted sentiment this is where I like struggled a little bit to understand. But correct me if I'm wrong, I think this feature here allows you to figure out the overall so instead of figuring out the overall feeling or emotions of whatever the text was, it pinpoints different feelings at certain points in the text. So it gets a little harder here like this is a very very Yeah, like very straightforward sentence, but I guess I can say like something awesome. Hey, real quick.

Brandon Minnick 34:59

Send to you You boss.

Diana Pham 35:02

So whatever we're saying about Brandon right here is positive.

Brandon Minnick 35:06

Oh, cool. So yeah, looks like I think would be cool. If we toggle between the entities tab and the targeted sentiment tab, we should see the same things underlying No. Okay. Well, close though.

Diana Pham 35:21

Sorry. So my name, your name, my social security number. And what's over here?

Brandon Minnick 35:28

Oh, your name my name. All right.

Diana Pham 35:31

Yeah, me saying that. It's my name. And then this is my actual name, I'm guessing. So

Brandon Minnick 35:38

looks like a targeted sentiment, but the sentiment analysis takes everything. So if you write 5000 words, have a look at all 5000 words and determine overall is that happy or sad? And here, it looks like it's breaking down the specific topics because you said Hi, my name is Diana. And here's my social security number. Then the next topic was we think Brandon's awesome. And so it's almost like it's found a different topics because parts of paragraphs can be positive or negative. And maybe we don't care about the whole thing. We're looking for a specific topic that they're discussing. How are they feeling exactly about? Brandon? How are they feeling exactly about Social Security? And it looks like he can break it down like that for you. Which is really cool. Yeah,

Diana Pham 36:28

yeah. Okay, well, pretty much those are all the Actually no, I think there's another feature I think it's like, there we go. Syntax. So it actually takes every single part of the text minus spaces, it seems, and labels the part of speech and how confident it is that

Brandon Minnick 36:48

percent confidence on some of those. Look at that. Yeah.

Diana Pham 36:51

Okay. Well, it's possible. I didn't realize it was possible. Yeah.

Brandon Minnick 36:56

Yeah, cuz it looks like this is breaking down every word. Is it a noun? Is it a verb? Is it punctuation? Is it an adjective? And yeah, I think like, it says, the period, the full stop is punctuation is 100%. Confident. And yeah, yeah, that'd be something we could be 100% confident in.

Diana Pham 37:17

I'm curious why this one is like, over point nine, nine. And this one is like, very confirmed. That is that is is affirm. But that was an interesting, look at punctuation. Alrighty, so that is pretty much AWS comprehend of all the features. I'm curious what all of you think, are the best ones or what you would use this for. But I wanted to do something fun earlier in the podcast that you joined a little late. I told you all that I met Brandon, at KCDC. Like last month? Well, no, we saw each other last month when we met the year before, right? Yep. Yeah. And so I spoke at KCDC giving that how to talk to them and talk and people have the option to give feedback. I think it entered them into a raffle that they gave feedback. And so I'm going to pull up my feedback from last year. I'm going to tell you right now, I went through some character development after that, after after reading. And we're gonna you're gonna guess what each of them is, and the confidence percentage, if that's cool.

Brandon Minnick 38:24

Yeah, let's give it a shot.

Diana Pham 38:28

I like accidentally pull this up crying. No, but I really do. Like I really do appreciate whatever feedback I end up getting at conferences, because you don't know how well you're doing until you hear back from someone. And yeah, I would like to think that it's enough for people to come up to you say thank you, tell you, you know, their thoughts, feelings, or whatever. But there are some cases where some people don't feel comfortable doing that. And I totally don't blame them.

Brandon Minnick 38:56

And cultures to I've spoken at conferences all around the world. And yes, sometimes everybody will come up to you afterwards and say, Hey, that was great. Thanks so much. And then you get a slew of negative feedback from like, their written feedback, or however the conference is gathering it. Or spoke at places like, especially like the Dutch are very straightforward. They'll they don't mince words, they'll cut straight down to it and like, hey, yeah, this was good. This was bad. This was good. This is bad. And you're like, Okay, well, thanks. I appreciate the honesty.

Diana Pham 39:36

Once again, character development.

Brandon Minnick 39:38

Absolutely. Well, then, yeah, it's feedback that I'll take in. Well, first double check, like, because obviously, I never want to get anything wrong, or it's never my intent to lie about something. And so yeah, it's the best to circle back on those things. Or it's like, hey, well, you know, you said this, and you're not wrong, but it's technically a little bit more nuanced and you tried to figure out a better way to navigate that for maybe your next talk or your next video or just that continuous improvement is is always a good thing. Yeah, yeah.

Diana Pham 40:11

Okay, so with that, let's get started. Here we go. All righty. So I actually full disclosure, I'm just copy and pasting. I have not read it. So I'm going to read it out loud to all of you. And Brandon's gonna guess. So. Let's go here. Thank you for bringing awareness to this issue. I am a woman and I think male attendees need more specific examples of what constitutes creepy behavior saying don't be creepy isn't enough. Many men have no idea what their behavior when that their behavior is creepy and unwanted. Because they have been allowed to behave that way their entire lives. In parentheses, women have been socialized to put up with and I actually can't confirm that that statement is true. Because women who have come up to me after this talk, we're like, yeah, at one point, I had a co worker who snapped like my bra strap saying that something showing and she did not flinch. Like cross her mind that that was not okay, until another male coworker saw that and was like, Did you realize that he like? There was no real Mal intent, but it was more like, Oh, something showing

Brandon Minnick 41:23

up? Yeah, yeah.

Diana Pham 41:25

So in my examples, like, I think I pulled up a few, but I don't think it cause for sure exactly what they said. Whatever I said was definitely not specific enough. And so when I kind of tailored that talk, and made sure to be more specific on what that is. So let's analyze this. And you tell me what percentage of anything you think this is. Yeah,

Brandon Minnick 41:47

this one's tough, because she is giving some negative or sharing a negative story with us, you know, and I agree. Yeah. And you know, if somebody's has only lived in one world has only been acting a certain way, then that's the way you think the world works. And that's just normal. And to say, don't be creepy. You're like, Yeah, well, I'm not. I'm not a serial killer. That's what's crazy. No other things can be creepy, too. But yeah, we're here. Yeah, it starts out very positive. So I definitely think if we go to the targeted sentiment, that that first sentence that thank you for bringing awareness to the issue, and I, that should be positive, certainly. But goodness, yeah, the rest of it. You know, to me, it's not negative, like it is facts. But yeah, I'm curious how the model would score it because it's not a pleasant topic. So, okay, here's, here's my guess I'm gonna go with the overall sentiment. I'll say slightly positive, maybe like a with like a point six, six confidence intervals, two thirds of the way there. But I think if we go to targeted sentiment, we'll see. Hopefully, I'm right on this. See a more clear breakdown of what's positive and what's negative, but let's run it and find out.

Diana Pham 43:24

So analyze, let's go over to sentiment. So here we have, okay, okay, I forgot about the mix. That's not fair to you. But neutral. Yeah. 30%. I am a woman. There we go. That's pretty neutral. Women have been socialized to put up with it. That's what I'm getting as neutral. But overall confidence. There's some amount of like negativity, because, you know, I would I was I would, I would guess it's the don't be creepy. They were creepy is kind of just negative in general. Yeah. And mixed. I'm not sure what mixed means. But that's there.

Brandon Minnick 44:02

Yeah. It's positive and negative at the same time. Yeah. a backhanded compliment, maybe.

Diana Pham 44:09

Perhaps. So the funny thing is, um, normally sentiment analysis models can't monitor or identify sarcasm. So it usually goes one way or the other. So I'm kind of curious what mixed would be but Okay, so that's the first one. Let's go into this next one. All right.

Brandon Minnick 44:31

I do targeted sentiment.

Diana Pham 44:34

I'm sorry. Oh, yeah, of course. Sorry. Targeted sentiment over here so you i Women, male attendees, more. Many men, they're they they're women. Those are the words that are underlined. Yeah. So sorry, everyone who's just listening and I'm not actually reading these things out loud to you. Yeah, so.

Brandon Minnick 44:55

Okay. So you the part of the pair paragraph that was about you is how to present positive. Okay, cool. I totally agree with that. Yeah, the rest is actually they scored it as neutral. Yeah,

Diana Pham 45:10

it's neutral. Yep. Okay. So let's drop this next one. And I think I sat there and cried for a little bit. And I was like, Get over yourself day and I just get better. Don't be sorry. Be better.

Brandon Minnick 45:25

No, that's, I guess great feedback. It's, it's unfortunate that this exists in the world. But certainly shining a light on it is one way to sunlight. What's the old phrase sunlight is the best disinfectant. So if we can shine a light on things like these talks about things like these difficult topics, and the conversations might be, I think that's how we move past this and highlight what is okay and what's not. Okay. Sharing experiences. Certainly. The best way to do it?

Diana Pham 46:02

Yeah. All right. Well, here we go. This big blur? Oh, no.

Brandon Minnick 46:05

Oh, yeah, I don't like that first sentence

Diana Pham 46:08

was very disappointing. As a woman in tech, I was excited to see this topic and encourage my teammates to go. However, I was embarrassed by the presentation. Her examples were of blatantly gross Twitter troll messages. So her explanations were just this is bad, or just this is bad. Don't do this. It's giving me the IQ. Instead of providing examples of real slides in the office, Slack messages, LinkedIn, and giving real data explanations and tips for interacting with coworkers on a team. Reviewing her LinkedIn, it seems she's very early in her career, and perhaps didn't have the experience yet. To be giving a talk this looks important compared to the rest. Compared to the great quality of the rest of the talks, I watched this year's and last year's conference, this one fell below par. So, so it's funny that they mentioned like, you know, her examples were blatantly, blatantly gross Twitter troll messages. The funny thing is, a lot of my examples are actually from women who have been working in the industry far longer than me. I told them though, I was like, I'm not going to disclose names just for their own privacy. But that was the case. And so I thought it was pretty ironic that they were like, she hasn't been harassed enough. And I was like, Little did you know?

Brandon Minnick 47:25

Yeah. Yeah. It's, it's, it's an interesting, it's interesting feedback. I I'll say from experience, well, not that I have any experience this topic, but yeah, like when I read the YouTube comments on videos, I post, I was trying to read between the lines, and here. It seems a little silly, because yeah, she's like, she didn't give any real examples. She only gave example. It's like, she didn't give examples. And she did give examples where that weird phrasing but yeah, it looks like she was looking for more like work based conversations instead of something that some troll would put on Twitter. I guess she's assuming that that same person who would say this on Twitter wouldn't say it at work. I don't know if I can agree with that. I think you are who you are on the internet. And the more anonymous you are, the more honest you are about yourself. So I think without without seeing your talk, it sounds like you're doing doing it right. But perhaps she's been or she's gotten messages on Slack and LinkedIn, it was looking for more examples on how to deal specifically with those. But goodness, this so sentiment analysis,

Diana Pham 48:48

analysis lies.

Brandon Minnick 48:49

I mean, she she uses the word excited. But then she follows that up with a however, so if the engine is good enough, but yeah, it shows pivot. Yeah, so there's a little bit of positivity in here. And I'm going to machine learn myself and go off of what we learned from the last score, where we saw there was some positive some negative some neutral and I'll say maybe, yeah, 10% Positive. It does feel very negative. But I don't know if it's enough to get past neutral. So let's say 10%, positive 20% negative and the rest neutral.

Diana Pham 49:36

All right, so Oh, sorry. This is the targeted sentiment. Okay. Yeah, no, you're pretty good. Like a lot of it was like a lot of it was neutral course there. Is that like

Brandon Minnick 49:52

she you're the she? Oh,

Diana Pham 49:56

yeah, no, that makes sense. Yeah, they got pretty LinkedIn, LinkedIn is neutral, for sure. Okay, cool. Let's look over at our overall sentiment. Lips. It was very, it was very nice.

Brandon Minnick 50:15

99% confidence on negative I was. I mean, I don't disagree with that. Yeah, I am surprised that the model she is ripping into her. Gosh.

Diana Pham 50:35

Um, okay, well, we won't look at this one. But one of them says the use was a little distracting. So at the end of my talk, I played the ukulele. And so yeah, anyhow. Well, some people are really nice, very strong speaker. Let's see.

Brandon Minnick 50:59

Yeah, you just got to stand until the next show comes on. So everybody, make sure you stick around. But we'll be cut off if we don't if we don't finish on time. So you won't see us at the top of the error anymore, but definitely stick around. Next shows great. You don't want to miss it. But I think one more quick one. Yes.

Diana Pham 51:19

Okay. Let's do this really short one. Oh, where is my screen, huh? Okay. So one last quick one. Let's look at this right here. This talk was all caps everything. exclamation point, exclamation point, exclamation point. I literally told the new girl in my department not to dress cute or she would be harassed all day. I'm so glad she's someone is shedding light on this topic. I was like, No, that is not what I meant. Um, yeah. But anyways. So I'm just gonna assume that this is really positive. Sorry, that probably wasn't the one I was like, trying to find. But that was the one where I was like, No, that's definitely not what I was trying to communicate.

Brandon Minnick 52:04

But also tough, right? Because they're complimenting the talk. But then I assume words like harassment will get flagged as negative, or I hope they would. I have no idea anymore. I was trying to go off the first example and the second examples answer and now I'm off the rails. i Let's see. Let's see. 5050. It feels after this is positive half of this is negative.

Diana Pham 52:36

Okay, let's not an actual positive or like a actual. Okay. Neutral, positive, negative, a little bit of everything cool.

Brandon Minnick 52:47

But mostly, mostly positive. That's like 71%. Yeah. Okay. All right. Cool.

Diana Pham 52:52

And then let's do one super fast. Last one. Oh, copy and paste. Wonderful content. It is unfortunate. We women live this every day in some workplaces. Having this content at a convention like AC DC can only help show this problem. Thanks, Diana.

Brandon Minnick 53:14

I'm gonna say probably the same results. Yeah. 75%. Confident. Oh, that's a word unfortunate in there.

Diana Pham 53:22

It's mixed. I don't know why it's. Oh, yeah. Unfortunate. Definitely mixed. Huh.

Brandon Minnick 53:27

Yeah. It's like positive sentiment about a negative topic. Yeah. Maybe that's what mix means. Maybe we just decided, well, do we only have two minutes left? Oh, good. Thank you. Thank you so much for for joining us today. For people who want to join us follow the conversation. Where can they find you online?

Diana Pham 53:50

Find me at Diana's oyster on LinkedIn. And I'll just continue posting there. I'm also sorry, I as I saw on Twitter, and I'm also on LinkedIn.

Brandon Minnick 53:57

Absolutely. Well, thanks again for joining us. Thanks for showing us around ATS comprehend today. And thank you for for watching subscribing, joining us for another episode of the dotnet on AWS show. Don't forget we live streaming on Twitch every other Monday. And we also release it as an audio podcast. So if you want to listen to us, let us join you in your balls, your buds. Your buds as you're walking the dog, subscribe to the dotnet on AWS show wherever you find your podcasts. We'll be there. And we'll see you again in two weeks with another episode of the dotnet on the dotnet on AWS show