Thank you to all the participants for building amazing projects in the Turn Language into Action Hackathon! It was great to see all of your projects and the ways to improve the world. Winners have been announced during this special NLP Livestream.
Hey everybody and welcome to the NLP livestream. I’m Brian Munz, your host for this week and for most weeks. But this week is going be very fun because our hackathon is officially over. I’m going to show off some of the winning projects and so hopefully it’ll be fun just to see all of the interesting things that everyone was able to do.
First off though, let me share my screen because I want to use a presentation to guide us through this. Ready, go. Okay. All right. I accidentally gave you a preview.
Before I get too far into showing everything or announcing the winners, things like that, I just wanted to give people an overview of the hackathon, maybe you haven’t heard of it, you weren’t here for the beginning of it, so just giving some background so that everyone knows what they’re seeing, things like that.
First of all, overall though, we were very happy with how many people joined. We had 530 people register with 46 projects. As you might know, the purpose of the hackathon was to create a web application, and the purpose with this was to use NLP in particular, of course, the expert.ai API in a web application project that had some a good meaning behind it so that the purpose for this was to do something that’s positive in the world, whether it be affecting change, enabling people to better understand, avoid disinformation, for example, to understand the impact on environment, keeping people honest, preventing hate speech, et cetera. That was the overall purpose. Generally, I can say that, very impressed with the creativity as well as the quality of a lot of the projects. You never know what you’re going to get, or how creative people are going to be. I can say out of the 46 projects, we were very impressed.
Actually in our comments, I’m going to put a link to the project gallery, and so I’d encourage everybody to go to that project gallery and just start clicking around. A lot of them have live working examples of their apps. All of the ones I show here today except for one, because it’s a bit more technical, but they all do have videos explaining the project itself. So definitely fun to just go through there and see what people did, but also the different ways that NLP can be used in a project with a goal like this.
So just as a reminder, the judging criteria was three things and they were equally weighted. One was the use of the APIs. The REST API was very easy to use. What we mean here is two things. One, is not just if you’re able to get and retrieve the data that we provide that our APIs provide, but also how that was integrated into the application. Where there creative ways that you then took that, and from a technical perspective changed the web application?
The second thing was creativity. A really awesome thing about hackathons in general is you have one foot in the real world and one foot in the clouds. And so this is the one where we’re talking about the cloud stuff, not the cloud, but your head in the cloud stuff of how creative you were. Maybe it wasn’t the most useful thing, but it was really interesting and cool. This will get you extra points for that.
And impact, what we basically mean by that is just overall wow factor. You’ll see later a lot of these projects were surprising in their use case or surprising, and it could just be how well the user interface worked, how well it looked, just overall surprise of, “Wow, this is a really cool and interesting idea.” In fact, one of the first ones I’ll show has that creativity and impact.
And of course something that is going to be important to everyone that you see today, is the prizes. So first place is $5,000, second place, 2,500, third place 1,000. And then also what we did is we had three categories that would have honorable mentions. And so those were focused on particular, which is around the social governance, environmental impact of companies. It’s a big buzzword right now. Hate speech and sentiment, so these would be projects that addressed those particular things in the best way. And so also if you had won first, second, or third place, then it was not really eligible for the honorable mention, because you’ll see that the top three winners applied to one of these categories of course.
So again, congratulations everybody. We were really impressed. Hopefully everyone had fun in building the project. And even if you’re not one of the winners, there’s projects that we found particularly interesting. We may reach out to you, it could be for a live stream, maybe for a blog post. But I would encourage anyone, it could be a way for you to gain experience and thought leadership, to get exposure to yourself or your company. We’ll definitely be reaching out to a multitude of teams, not just the winners.
Without further ado, what I’m going to do is go backwards. So honorable mentions first, then third, second, first to announce the overall winners. If my presentation … Well, I’ll click. Here’s the category winners. These are the honorable mentions I mentioned.
The first one is Hate Speech. And so this is one that doesn’t make sense for me to demo really because it is more of a tool that runs behind the scenes. The team was a open source tool that was built by this team, which tries to provide investigators with a way to investigate war crimes and things like that. And so it’s a larger application, but where they use expert.ai is to analyze the text. This could be for example, just communications that they’ve found, social media postings, things like that, where it’s to analyze these posts for different aspects of whether it be hate speech, to better structure this unstructured data within their application. This one I thought was especially interesting. The project is decentralized, so there’s transparency in the investigation with this project, which is meant to basically aid investigation of war crimes. So of course the for good part should be pretty clear there. So congratulations to citizen5.
Next is ESG. I’m going to show this one quickly. I apologize in advance to the teams if you think I’m flying through too fast. Most of you will get a chance to present this yourself if you want. I just wanted to make sure I’m sensitive to time. But this is SustainaMeter, and basically it’s meant to provide analysis of companies and their attitudes towards ESG, and then have that be comparative to other organizations. This one off now. I may be jumping back and forth, so hopefully it’s not going to quite be as clunky every time. This is the SustainaMeter. Yeah.
What you can see here is they’ve provided some examples here, which is nice. It has LinkedIn content analysis. These are posts in LinkedIn, by these companies here. And so let me pick one which may have had some controversy.
Exelon is energy and sometimes energy means environmental issues. And so you can see here and analyze the posts. It comes down here and it tells you what ESG categories it found. Environment, of course, that makes sense. It’s an energy company, and it tells you the different subcategories in environment, waste and emissions management, environmental opportunities, and then their total ESG score. But then it also tells you whether it’s positive or negative. This to me looks like since it’s positive, they are trying to show themselves as being changing for good in this way. Also, it has a variety of other charts down here. But this is a interesting thing that you can imagine being used for of course, ESG analysis and it had a very nice user interface. So this was a very, very cool project I think. Also, I heard you can add your own posts in here, evaluate them based on their criteria and the analysis that they built. You can of course change the parameters of the search. Back to the presentation. This is not going to work if I’m going to keep showing off the winners.
The last of the new category winners is Sentiment & Emotions category. So of course this is how projects used sentiment analysis as well as … Expert.ai has emotional traits identification. In fact, almost all the projects use this aspect in some way, which makes sense because we’re talking about hackathon for good.
There was several projects that addressed news, and fake news as well as disinformation. This one was one that stood out, we thought it’s called News Report. This is supposed to be a report card for our news reports. And so what it tries to do is when you post an article … yeah, it’s better. It’ll evaluate that article for problems, whether it be disinformation, hate speech. It’ll also give a score of the source of the article. If it’s A, then the source is relatively reputable, the large media outlets all the way down to as low as you can get. I’m not sure where the bottom is. But here I’ll show you. I found an article that was particularly bad, that contained hate speech and things like that. And so you can see it has identified the article. I may have let the thing time out.
Okay. One second, because my thing timed out. The site timed out I think because it is… So let me find this horrible article. Please don’t read it. Okay, there we go. Sorry about that. The report card here comes up and it gave the news outlet a U, which is either worse than an F or means unacceptable. I’m not exactly sure, but means that it’s not the best source when it comes to fighting disinformation, or just hateful rhetoric. So you can see it gave emotional language a C. The idea of course being that the more emotional language you have in an article, chances are they’re not being unbiased. As well as it found hate speech, which of course is not going to be good.
But another interesting thing it found is it tries to identify the most hateful sentence, as well as the most emotionally charged sentence, which is also an interesting way to identify ratings. Let’s quickly get off of that horrible thing. But yeah, I thought this was an interesting project to use where people can actually put in articles, see the analysis. It had a nice user interface of course, and nice usage of IPS. Congratulations of course to all of our category winners.
I keep previewing the winners. But I should mention too that if you have any questions about the particular projects, now would probably not be the best time because they’re most likely going to present in a later live stream or in some capacity and I don’t want to answer technical questions, especially for them. So if there’s questions about the hackathon itself, feel free.
Third place is MeMo. I’ve been pronouncing that wrong the entire time. I’m guessing it’s MeMo now because that’s a actual word. This one was pretty interesting because it dealt with the metaverse. The idea here is that the creator of this project has a community where people are encouraged to share applications, games, things for VR, for metaverse. And what it wants to do is use NLP to moderate these entries. And so they use the NLP for classifying the different entries and to [inaudible 00:16:09], you can see here it does emotional traits. It also blocks PII, so personally identifiable information to protect the identity of the users. It puts it into a moderation queue so that the moderator can determine if it’s appropriate. Also, hate speech detection. Obviously you don’t want people making comments or putting games and things that are anything around that.
I will show what I can from this because obviously I’m not going to open up the metaverse and jump in there, but I will quickly show this video of what they’re talking about within the metaverse. This is within VR, submitting a game to this community. And so you can see the text here of people either naming the project, making comments, and then the analysis happening of, there you can see in fact, what the APIs found around the sexism, which is within hate speech sentiment.
But the other aspect to this was not only the VR metaverse landscape, but also moderation, which I can’t show. And so this is the user interface for the moderator of that metaverse area community. I’ll even refresh this queue. So this is where these different submissions would come in, and the person can see the submission. They can approve or reject it, they can see where this is reported for these hate speech items. This is blocked because of PII. So the person can go in, determine if this PII is appropriate to be there. This again was a very, very interesting and creative usage of NLP. Kudos to MeMo.
Moving on. Second place is atlatl who built the, Understand Human Rights Complaints submission. This is another interesting usage of our APIs, and an interesting project in general. Atlatl is a startup, they enable companies to have comments and submissions happen all up and down their supply chain, where someone in that supply chain can report. And I like a whistleblower or something, can report something that is happening, that is potentially of course at risk and damaging to the environment, or is abusive, or something, but to preserve anonymity of the person submitting it as well as… What they end up using NL API for, is to analyzing the complaint, identifying which of the ESG goals that violates for the company, understanding the emotions of the complaint, as well as detecting PII. If the user has entered any personal identifiable information about themselves, it protects the whistleblower from their name getting out there and they could potentially get fired, which is something that happens often.
What I’ll do is go over here and show… So what they’ve done in for this hackathon is they have a larger product, but they use the NL APIs simply for this part of it, the complaints and the analysis, the complaints. And so I’m going to show you quickly what that looks like. I’ll just use one of their examples. This is obviously a form where someone that works for this particular company would make a complaint. And so you can see here it says, “Do not tell my boss that I complained. I’m afraid he’ll fire me.” Says where they work and then begins to talk about pollution that the factory has done and that the workers are under bad conditions there. And it seems that they are also being yelled at.
But there’s a lot of aspects to this. But you can see it identifies the emotional traits within the complaint, of anger, their fear and shame, which I think is pretty accurate there, as well as ESG categories. It accurately showed that it was in the environment, it was for biodiversity and environmental crime, which is interesting because it does seem to be something that is happening on a large level, it could be a crime. And then if it finds PII, it redacts it within the comment. So this is what, of course, the person that is moderating and taking these comments, what they would see.
This I thought was very interesting and you can see the positive impact here of enabling whistleblowing and enabling people to safely show when their companies are not on the up and up. I think it was very, very well implemented also.
I just wanted to show an example too. You can see also it’s better for PII where you can see the names of coworkers was redacted. So that’s important because that’s somebody who could be targeted for some repercussions at their workplace. So very cool project again and definitely I keep saying it, but go to the project gallery to check all of these out. There’s much more detail in there. And also cool projects. Several of these projects were startups, so very interesting to see what people are doing.
Without further ado, first place is PagePal. This one we thought was extremely creative. It was a use case I hadn’t ever thought of before with emotional traits recognition especially. But so if you know anyone like I do in your life who suffers from autism spectrum disorder, one of the big problems with that is their ability, especially when they’re young, because part of it is learning cues. Learning how to understand what other people think, how other people work. And that’s the way that kids with autism will learn how to navigate the world.
And so the really interesting use case that PagePal came up with was to use NLP to identify the emotions within text as a way for kids to teach themselves how to identify emotions. So for example, especially in this case, they use it for children’s stories. Because for example, in a story it may imply that a character is angry, or sad, or upset, but the child may not understand. And so what the PagePal does is, points this out to the child, so as they’re reading, they can learn. And so again, go check this site out. But what they do here is they provide some books that are analyzed for these emotional traits. I’ll just pick one randomly.
You can see here it identified suffering and you can see why it actually came up. “Uncle Wiggly felt the sharp pain, stood still for a moment.” Affection, so this is one in particular I think is good because they are not specifically saying we are very sad. It is saying we are sorry to have you go. And so for a child with autism, this is helping them understand this part of the text is expressing affection, and this one is expressing happiness then.
So yeah, I loved this project. I think it was very interesting use case, like I said, I’d never seen before. Similar to all of the projects, I can see how these could be used in the real world. I hope not only for this project, I hope for all of the projects that this work continues, because I think there’s a lot of really good ideas and a lot of really good implementations and use cases. So I really look forward to seeing if and when a lot of these projects continue.
What’s next for all of these teams? Again, congratulations everyone. And for anyone who didn’t make this top six, it was very competitive. There was a lot of really good projects. All of them had something that were impressive and had an interesting way of using the APIs. And so we definitely appreciate everyone who contributed, everyone who participated. It was a lot of fun seeing what everyone came up with.
But one thing is that we’re going to do in the future is we want to dive into some of these projects to see what kind of interesting things that they discovered, things they did while building the projects. So future live streams, keep an eye out for some of these teams. We may do some run throughs and maybe some Q and A with them. But we want to have them, the ability to get them out here to see what they did.
Also, we’ll be releasing blog posts on the winning entries, perhaps some more. And there’s going to be some promotion with a press release, which I believe may have already come out today. And then of course the prizes. All of the winning teams will be getting emails relevant to that.
But like I said, again, this was really fun and I hope you enjoyed seeing all the projects. We had a blast looking at them all. And so yeah, that’s a wrap on our hackathon. We hope you enjoyed this. Oh, actually next week is Thanksgiving, so I don’t think we will be showing next week. But you’ll see us in a few weeks with as usual topics relevant to natural language processing.
And again, look for us in the future and have a good rest of your day.