[MUSIC] In our last video, we saw how fast it is to extract big data from anywhere in the world and get it ready for analysis. In this video, we're going to take the next step of looking at how that big data is transformed into big information to help you with your marketing programs. Our guest is Mr. Seth Redmore of Lexalytics. I found Seth by asking my professional network who would be the best to teach me about machine learning and text analytics. When a number of them said Seth Redmore, I contacted him. This is what we want you to accomplish with your professional persona and in building your professional network. Welcome Seth. >> Thanks Randy. >> Can you tell me a little bit about yourself and Lexalytics? >> Sure. Right now, I am Chief Marketing Officer at Lexalytics. I've been in tech for about 20 years. I spent about 13 years in networking technology, culminating with founding and selling a company to Cisco. Then while I was at Cisco, I started to get really interested in how can we measure the effectiveness of marketing beyond things like revenue and stuff like that. How can we look at how we're changing the conversation and so, I decided to start looking into text binding and text analytics. And so, I've been with Lexalytics for six years. Lexalytics has been around since 2003. We shipped our first product in 2004, which was the worlds first commercial sentiment analysis engine. Now we are the strongest player in social listing, social marketing applications, and customer experience management. Our customers process about 3 billion to 4 billion documents a day through our system. >> Wow, that's amazing. In this video, we're gonna be talking about text analytics. If I posted something, say, on what sort of information can you extract out of there, from reading that? >> Sure. So the way to sum it up is you give me a piece of text and I will tell you who, what, when, where, some about why, but most of it so you can figure out why and understand what you need to do to change your business and to make more money. There's a couple technical terms. So there's something called named entity extraction. That's the who, and the where. It's company's products, places, people. You can make other things entities, but those are generally what they are proper nouns. And you make them and entity, so you can associate other things with them, like sentiment. Themes are automated topic extractions. What's the buzz? What are people talking about with this? Are they talking about the comfy chairs in there? Are they talking about the cold pool? Are they talking about the hard bed? What are the things that are really coming up inside of the conversation? Categories are the other side of the coin from topics and themes. So categories are things that you are interested in following and you know you're interested in following. So you tell the system, okay, I want the following categories. Location, service, bedroom. Those sorts of thing. Intentions is a relatively new technology. So it's what is someone going to do in the future? Are they going to buy something? Are they going to sell something? Are they quitting your service and buying somebody's elses? So intention analysis allows companies to reduce turn, find new customer leads, and then overlay ed over, sorry go ahead. >> Well what would be an example of intention? >> Yeah, so oops, I dropped my camera. Now I need to buy a new one. Very simple, very straightforward example. There's a lot more sophisticated examples than that, but once either a camera or a review site or a camera company sees that, that's a great time to reach out to them with a tweet and say hey how about 20% off on your next camera. >> That's an example of real time marketing. >> Exactly. So sentiment analysis is layered on top of this. You can get sentiment analysis at a number of different levels. And it's really important to have very fine grained sentiment analysis. So the sentiment of my company, inside of this article, the sentiment of another company may be completely different. And if you're only looking at document sentiment of the whole thing, you're gonna miss out on some of those differences between there. >> You talk about sentiment. What does the term sentiment mean? >> Sure, it's, are the opinions expressed inside of here positive, negative, or neutral? As related to the entity, or to the theme, or to the overall category? Because, for example, a category, like energy exploration may be somewhat negative in an article, but there's a company that's doing really well inside of this. So, they'll be positive in that article. So, it's important to tease out where you're being seen negatively in a positive place and vice versa, so that you can take advantage of that. >> Okay, but what about if I'm using something like sarcasm or attempting to tell a joke, or maybe there's an oil spill, and I say it's the biggest or the most colossal, which sounds positive but it isn't. How do you handle that? >> Sarcasm and humor we handle perfectly. That's a different issue than the biggest and largest. So the contextual understanding is completely important for understanding whether a word is a positive thing or not. From the perspective of oil spills everywhere, being a biggest oil spill, that's great. Oil spills love that. But from the perspective of the Gulf States that were affected by, say, a very large oil spill, that's a negative thing. So, it's all about context. You need semantics, what could the words mean? You need syntax, what's the order of the words, and the structure of them? And then you need context to determine whether something is positive or negative. And even to try and figure out whether it's humor or sarcasm, you need to know, contextually speaking, what did this person talk about before that suddenly they're talking about this very differently. So if they're really positive on something they used to be really negative on, like even just the tweet before maybe, maybe, maybe just a hint of sarcasm. [MUSIC]