We stand with Ukraine

NLP Stream: ESG, Making Decisions in a Woke World

More and more, people are making decisions as much with their hearts as they are their minds. Be it a consumer product or business investment, purchase decisions are no longer based solely upon the cost and quality of a product or service. Instead, buyers are weighing environmental, social and governance (ESG) criteria heavily into their final decisions.

In this livestream, Brian Munz dives into the ideology behind ESG, why it’s become an increasingly large part of business decisions and how NLP can be used to identify ESG concepts within text to enable faster analysis on large volumes of unstructured text.

Transcript:

Brian Munz:

Hi, everyone. And welcome to the NLP stream, which is a weekly series dedicated to the latest and greatest in natural language processing and natural language understanding. I’m your host, Brian Munz. I am the product manager and developer advocate here at Expert.ai, who are the leaders in NLP technology. So what we do here every week, it’s kind of a meetup style. So we have some things that are going to be more technical, we have some things that are going to be more broad, but generally we want to highlight things in the NLP world, which are going to be of interest to anyone interested in NLP. And so, things may be lose some weeks, may be more polished other weeks, but in that way, if anyone has anything they’d like to present on in the NLP world, you can contact us. You’ll be able to see that later.

Brian Munz:

But we welcome everyone. We want it to be as communal as possible and to basically feature things that are going to be of interest to you. So this week we have a special guest, someone who I don’t always get along with, but I’ve spent a lot of time with who is myself. And so today I’m going to be talking about… Oh, let me share my screen. I’m going to be about the concept of ESG. And so that is something that has become more and more of a topic in the world of finance, especially. And to note, today I’m going to talk about ESG and how people are making decisions in a woke world. And so just on that word itself, woke, I mean, that’s one that has become somewhat charged and an annoyance to some people in the business world, but at its core, it’s a word… The original meaning of the word basically is to become awake to social injustices in the world.

Brian Munz:

And so through the internet, through the world becoming more flat, through access to information, let me move this off, access to information, maybe better access to education and more educated populace in general, people have become more aware of injustices in the world and issues that are facing the world in general, whether it be our planet, the people on the planet, corruption, fraud, et cetera. And so, basically at that root of everything, I think it’s hard to argue that the rise of a socially conscious consumer and investment, it’s hard to argue that’s a bad thing. And I’ll get a little bit into the different factors around ESG in that way.

Brian Munz:

But generally, some stats today is 64% of consumers support what they feel are socially conscious brands, which is definitely been on the rise. And that poll was done during COVID. And so it’s even during times when times are tight, it’s something that is becoming more and more important to people and about a third of those people say they plan to increase that type of spending and not surprisingly also 80% of consumers like to spend money on brands that they trust. And in fact, I believe it’s about that same number will spend more money on brands that align with their values than ones that may not. And so this is obviously a concern for any businesses because very business can be affected. It could be as large as McDonald’s where even people who aren’t looking to spend on just brands that you’d find in a whole foods, people have other options that they can use, and they will go to other people when they find it.

Brian Munz:

And so there’s a lot of information going around. And so it’s within a lot of business’ best interests to monitor and keep an eye on their perceived effect on the world and the people. And so this kind of thing also isn’t new, right? Boycotts are one of the oldest ways that consumers have used business to convey their feelings about their practices. So by hurting a business at their core is to make money, boycotts have always been effective throughout history. And so this is just basically an extension of that, but a little bit more sophisticated and maybe more difficult for a business to, of course, monitor because boycotts are very overt. But yeah, so even if a person is kind of skeptical or think it’s a bunch of hippies run wild and with their peace and love stuff, I’m going to hopefully show that it doesn’t really matter because it’s in everybody’s best interest.

Brian Munz:

So moving on to what ESG actually is, if I can. Okay, there we go. So what is ESG? ESG is something, of course, that’s been on the rise within the last five to ten years, and so much so that it’s a bit of a buzzword and has been factored into a lot of investing practices. So the E in ESG stands for environmental. So that’s of course a business’s impact on the environment in the world, sustainability, carbon footprint, et cetera. Social. This is how the business deals with treatment of workers, diversity, support of minorities, discrimination, general scandal. A good example of this of course, is Tesla, Elon Musk, who just today called ESG a scam, I think, and I’m going to show an article later, which may, I didn’t plan this, but may be a reason that he’s upset about it. And then the G is governance, so governmental.

Brian Munz:

Believe that’s supposed to be governance, but what this means is basically the way a business conducts its business. So is it using ethical and transparent business practices? So this could be of course, fraud or accounting nonsense that they may use to hide failures in their business. And so these three factors generally are how a business affects the world around them. And as I’ll show, how it is in the best interest of anyone who is investing or wanting to purchase from a business whether they should give their money to them. And so what I want to do here is compare the two different views around ESG. One is the business view. And so why this is good business versus the personal kind of social justice view. And just to show kind of how they… It’s not just one or the other.

Brian Munz:

And so one reason to pay close attention to ESG is that chaos and uncertainty just generally, it’s one of the laws of business basically, for the markets of… Scandal and things like that causes uncertainty in the business world, so it can cause stock to drop, it can cause people to lose faith in the business and the leadership. And so it’s something that through monitoring ESG, it’s something to keep an eye on in terms of what may be coming and what to think about in terms of risk before an investment. One of the more personal type benefits is it gives consumers a voice. And so it’s making sure that a consumer by evaluating these things will know that their money is going to perhaps a better option, may not be perfect, but the better option.

Brian Munz:

Another way that a businesses use ESG is reputation monitoring. So besides just investment, if you are a company like any company, for example, Tesla, you would probably want to keep an eye on what people are saying about, you monitor the news feeds, monitor social media, and that way get ahead of any upcoming kind of scandals or especially if they’re not true, to prepare for, to engage with things that may be coming at you. And it’ll give you a good idea of the general goodwill that your company does or doesn’t have. And again, I mean, that’s who, I should have mentioned is this isn’t all about negative, of course. So within ESG, it’s also a matter of you can find businesses that have a very positive view amongst their consumers. Like in the environment world, you have ones like Patagonia and things like that, which are highly embraced by a sort of outdoors community.

Brian Munz:

And so it’s also a matter of monitoring your representation. So if you had had a marketing campaign, for example, that where you said, “We want to really appeal to outdoorsy types, where we want to be seen as an active lifestyle,” you can also potentially see the effect of that campaign through monitoring communications and articles using ESG concepts. Another kind of a good thing on the other side is it rewards good behavior and punishes bad behavior. So of course we have regulations and there are laws, things like that, but as we know, businesses are very good at avoiding some of these things. And even if they’re not interested in the thoughts of what your average person thinks, if this is something that can hurt them. So just generally, it’s good to keep businesses accountable and have them understand that there’s, there’s more than just abiding by the law, there’s abiding by your consumers.

Brian Munz:

And I’m kind of in that way, on the business side, it can help you stay ahead of regulations. So one way ESG can be used if you’re investing is if you know there’s some upcoming regulations in a certain country around the environment, around waste and sustainability, if you are doing your research, then you are able to determine, oh, this company generally is not prepared or has a bad history of a sustainability and is not ready. So that might make you think twice before you are investing in it, because whether anyone likes it or not regulations are coming down the road. And so that’s another thing to think about. And on this side, it’s less compromising of your beliefs. That kind of goes without saying, but having that a better feeling about what you’re doing.

Brian Munz:

Another thing that’s become interesting and important on ESG, which I know has been felt by companies like Facebook, things like that, is if you start to gain a bad reputation, it can be harder to attract talent. The morale in your company goes down, of course, and that will eventually affect the bottom line. And so even if you’re not overly interested in that, it’s something to consider that with this kind of new world of easier access to information and again, that could be for better or for worse because in ESG there’s also a lot of misinformation out there, but it’s something again, so you monitor and get ahead of. And then of course, there’s a brighter future for mother earth and humankind far out. Yeah. And so basically what I’m trying to say here is that it’s something that hippies and businessmen can finally agree on and that it’s essentially better for everyone. It’s one of those nice sweet spots where doing the right thing and encouraging the right thing actually is good for business.

Brian Munz:

So to get to the NLP side of things, because why am I talking about this? You can find out about it anywhere. NLP is something that is very, in my opinion, primed for helping with ESG. When you go out and you look for solutions in the ESG space, almost always it’s a platform where you can, it’s a database of sorts, you can search across a company. You retrieve a series of articles, generally a score, positive, negative score in the various categories of ESG. And so it can be… But the issue is that it’s a very complex topic because it’s very subjective. There’s obviously a moving bar of which issues people are interested in, which is most important to them. It may be regional for example. And so these systems are good, but where NLP can help is by providing a way to customize these solutions to yourself.

Brian Munz:

And so, again, it’s very subjective and there’s always going to be a human element in terms of, for example, if you have an article stream and it is identified through NLP that this article is overly negative about a company in the environmental space. That article may not even, may be misinformation. It may have slipped through and may be bad. So you’re going to need some sort of a human element in a way to monitor those things. And also again, you may not be interested in some of these concepts. And so with Expert, because we are an NLP AI company, we focus more on building solutions, which can be integrated into other projects, for example. So if you are building your own custom ESG research tool, we have models that can be used to identify ESG concepts and things like that.

Brian Munz:

But generally, something I wanted to point out is that the concept of ESG is very close to the concept of sentiment analysis within NLP, which is to say sentiment of course, is somewhat subjective. It’s built around algorithms and it’s built around the best efforts of developers in the NLP and AI space to best identify sentiment. And in that same way, ESG can be very complex. Sentiment, for example, will usually take the full block of text and apply a score positive or negative to it. And so one of the complications with ESG and with sentiment is that there may be an article that says, “Company X dumped 4 trillion pounds of trash into the ocean,” but then the rest of the article says how good of a job everyone else is doing. And so will that article be defined as a positive or negative overall ESG score?

Brian Munz:

And this is where some of the complexity comes in, where in building a customized system and having an element of oversight is going to be useful. And so what I’m going to show today is a demo in a minute, is at Expert we built a knowledge model, which is basically like a model that you may know within NLP and within AI, where typically it’s with machine learning. With Expert, we use a symbolic approach to NLP, which means it’s more rules based, which also means you can open it up and it’s more explainable, where with machine learning, you train a model and then kind of let it go. It’s a sort of a black box. We have hybrid capabilities, but we also focus quite a bit on symbolic approach because it means that you can more easily customize it.

Brian Munz:

And so that goes back to what I was saying before, where what we’re trying to do is create a baseline for ESG analysis and then using our platform, you could customize this analysis to exactly what you’re looking for. And, and again, if you wanted to focus on specific industries, companies, et cetera. And so what we tried to do within this model is there’s several facets to approaching ESG with NLP. The first, and these are pretty standard concepts of an NLP, but first we classify text to ESG concepts where what that basically means is if there’s a block of text, we have a taxonomy that it tries to apply that block of text to. I’ll actually switch over here to show you.

Brian Munz:

So what this basically means is there’s a block of text. It’s pretty much just saying, is this block of text related? Which of these classifications is this block of text related to? So again, it could be environment. And to the one I mentioned before could say negative, waste and emissions management. And so all that’s doing is it’s not really applying any kind of subjective score or anything onto it. It’s just saying we’ve identified this block as being relevant to, and it’s of course, negative. So that is sort of within there, but it’s related to this particular topic. And so that can be useful. Of course, if you’re building a system where you want to search or find different topics around different entities.

Brian Munz:

Something else that would, of course be just as more useful is to once this text is classified within that text is important data. So the extraction is to extract metadata from the text and I’ll show this in the demo, but basically the idea is, okay, you can imagine if it’s identified that a certain company is going to lower its carbon emissions. So rather than just saying that it has a positive impact, which is waste and emissions management, an extraction will try to give an idea of the scope of that. So it’s not just that it is bettering emissions, what kind it might be CO2 emissions, that kind of a thing. So there’s data that’s useful within that text that could be useful in analysis as well. Excuse me.

Brian Munz:

And then lastly, it’s around basically analyzing all of these, these things together, which is to derive sentiment. And so what, what that basically means is you can have the text that’s identified and how negative or positive that is to an ESG concept. And so you can see here within the text, it was identified as aiming to increase oil production, which is generally going to be negative on the environmental side. It would apply a score to that. This is something that it found kind of slightly negative, and then down here, it has some positive ideas where it’s moving to low carbon solutions, et cetera. And so in that way, what this model will do is it will apply scores to this and how the positive, or how negative this is give an overall score to the text itself. And I’ll show a demo in a second here of that. But overall, those are the three things we’re trying to do and extract in the text. So I’m going to jump off to a demo really quickly. I’ll show the platform.

Brian Munz:

Let see we have the trippy here. See what that’s called, recursion. So I will sign into the platform because it looks like I was signed out, but this is the Expert.ai platform where you have the ability to create NLP models, to train them, to test them, et cetera. And so, as I mentioned before, the idea here with an ESG model is that what Expert has done is provide this framework that I sort of outlined, and you can test text against this framework, which is what I’ll do here and see how accurate it is to your needs and to the custom solution you’re looking to build. Excuse me. So one thing first too, is this is a developer centric implementation. And so once you’ve used this nice user experience in UI to build out your models and to run all of your tests and things like that, ultimately most often it’s going to be exposed through an API.

Brian Munz:

And so when I show the output of this, it’s going to be within JSON which, if you’re not a developer, may look messy, but if you are a developer, you probably would understand that. Programmatically, you would go through this text, and remove all of the relevant details. In the future we will have nice looking, fancy demos like we do on tri.expert.ai, but this is a fairly new knowledge model. So yeah, I’m going to show three articles, each one showing one of the concepts of ESG. So first of course is environment. This is an article that talks about Volvo cars and a fossil free steel ambitions of Volvo. So of course you’d expect this to be related to the environment. Let me just copy and paste that over here, because I tried to have it a bit easier to get. So I’ll just talk to fill up the space. Okay. So we’re going to run this model against this article and see what comes up.

Brian Munz:

Something to point out first is when you see categories here, this is what I was mentioning before in terms of classification. And so for example, here you can see that there’s a positive feedback, score is about 50 for environment, which I think is going to be kind of the trend here, although there’s negative also on the environment. Probably because it’s going to mention, and this is again, one of the complexities of NLP in general, because it mentions both positive and negative aspects to this. And negative would be that in the past, Volvo was not really good with this and then the positive would be that they are improving. And so this is something again where you’d want to build this into a sort of not algorithm, but a system that would do your best needs. So again, negative climate impact was also from most likely talking about the past. This is the classifications that I mentioned.

Brian Munz:

So extractions, this is where you will find different entities that it identified within the text. And so this will be… So within our core engine capabilities and in NLP, something you’ll almost always see come back is it will identify within text entities. And so this is things like SteelZero or Volvo. And so this is just useful in NLP in general, and down here, we also bring back the main phrases, the main sentences, useful things for understanding what text means, phrases, et cetera. This is just core Expert.ai, core NLP technology. And so you could use it… The stuff that is focused on, the history is of course up here. But within this extraction, again, we identify entities that are related to the ESG concepts. So ideally, it’s going to find things that are relevant to the concepts in there.

Brian Munz:

So you can see here, for example, it identified that in the text, there’s a… Maybe I’ll actually… Sorry if this was too small before, which it probably was, especially when I’m showing little code like this. So you can see here that it identified CO2 emissions cut, carbon neutrality, which is related to carbon neutrality and it’s relevant to Volvo cars. And so this is [inaudible 00:28:02] job in sort of identifying the concept and who it’s relevant to, what entity it’s relevant to. Also CO2 emissions cut, we found that again related to SteelZero. Another thing I wanted to show here that’s also interesting is this is what I mentioned before around sort of metadata. It identified a CO2 emissions increase of 7%. And so it doesn’t make a judgment call on that necessarily, but it knows that it has identified that concept within this text. So that’s going to be essentially important within ESG analysis as you’re integrating NLP into your solution.

Brian Munz:

And then lastly, I wanted to just show, you have our general sentiment here. So the environment performance tour within ESG, overall the value is five. Excuse me. And so while that may seem low again, what it is trying to do is determine the overall across this article. And so it has found that it’s overall positive, which is good because the article overall is about the promises that are being made to be more sustainable in the auto industry. Oh, geez, I’m going too long. So I’ll be faster from here on out, but here is, I’ll skip that one because that one’s not as interesting. So this is one around social and I am going to copy and paste this when I get to it, but it has disappeared from my document. Interesting.

Brian Munz:

So I’ll just copy and paste it from here. That was dumb. All right. So this is around social stuff. And this is an article that where Tesla was accused of racial discrimination in the workplace. Tesla has generally a lot of problems with workers’ rights, and that’s not a surprise that they would have some ESG concerns. And so again, this article is about some lawsuits that have been raised against Tesla in this space. So it gives you guys quite a few categories generally, I’m sure we’ll see mainly… Yeah. So this is interesting, human capital a negative. So it’s identified… Something I didn’t point out before either is it also identifies the positions within the text where the relevant text is identified this as being positive or negative issue. And so that can be of course, overly helpful when trying to connect the dots between which parties are overly negative and which overly positive within an article.

Brian Munz:

Another thing is a controversial profile. This I am assuming is, and I will see below, related to Elon Musk because he is a controversial person, as we know has caused problems in their stock just by being himself. So within the extractions, something I wanted to show quickly, because I’m running out of time, is overall it finds in sentiment, an overall score of negative 200. So that’s not surprising, but of course it can be across a variety of factors. And so you’d expect within… This is, let’s see, social performance score. So that’s where most of the negative is coming from. It’s negative 175 in social of course, because there’s… It’s just this article is packed with negative information about their practices of in this area of racial discrimination and that they’re having a lawsuit done on them.

Brian Munz:

So this is, again, an example of something that could come across, come through a feed of news, a feed of articles or social media, where of course it’s an extreme example, but smaller company, medium size company, it may be something you’re not aware of, or maybe it’s a pattern. There’s a lot of larger companies that have a pattern of this, where it will reduce your confidence in the management, those kinds of things. And yeah, so all I have to say, hopefully I have shown you kind of how NLP can fit in with addressing ESG in a company. It’s obviously a very good match because ESG is mainly based on communications and your outward perceptions through the world. And so using NLP and doing analysis on text is a very good way to do that.

Brian Munz:

And so yeah, thanks to everyone for coming in this episode. If you have any follow up questions, please leave them in the comments below and I’ll get back to you soon. If you enjoyed this episode, join us for next week for, it is called Three Unique Examples of Impactful NLP NLU Applications. So going to see some, some real world use cases of NLP and NLU, three examples. And then again, you can reach us on community checkout developer.expert.ai. You can try out some of our natural language APIs for free, and you can also see some of this stuff in action on try.expert.ai. So until then, I’m Brian Munz and thanks for joining me. I’ll see you next.

 

Related Reading