Watch how you can use the MindMeld API to add voice-driven intelligence to your websites, apps, and devices in minutes. Get started today at developer.expectlabs.com
Hi, I'm Tim Tuttle, the CEO of Expect Labs. Let me show you how easy it can be to voice-enable your apps, devices, and websites using the MindMeld API. The first thing you have to do is go to the MindMeld Developer Center which is located at developer.expectlabs.com. You can sign up for a free account and get started immediately. I have already signed into my account so I will go to this management console that you see after you log into your account. I will create a new application that knows everything about movies and can answer questions about movies, so I'm going to call it "Movie Genius."
The first thing you have to do when creating a new application is crawl a collection of content that is then used to create a knowledge graph that powers the intelligence part of the intelligent assistant. If that content is already on the web or on your own website, the easiest way to do that is using the Crawl Manager tool that you can launch from your Management Console.
So this is the Crawl Manager tool. To get this tool running, you simply need to point it to a domain, which is typically your website or other third-party website. In this case I will point it to a website that has a lot of great information on movies; Rotten Tomatoes. If you want, you can fine tune exactly what is used to construct this knowledge graph by adding a white list and a black list. You can read more documentation about how to set that up in the developer center. And then you click, "add."
Now what happens here is on the backend, the MindMeld platform is spooling up a number off processes, and what these processes do is go to the website that you indicated, and then analyze the website so they can build the knowledge graph from that content. The first thing they'll do is look for sitemap files to get an understanding of the site, schema.org tags or other semantic markup tags that might be on the page, Open Graph tags if they're on the page, and then they'll use the structure of the pages themselves to try to identify the unique objects of content that will go in the knowledge graph. So in this case, the crawler is looking for objects that are movies.
You can see, once the crawler understands the website, it will go from page to page to look for objects that match the patterns that indicate it's a movie. On the right-hand side, you can see those objects populated in the knowledge graph one by one as the crawler finds them. If you want, you can drill down and look at the data for one of these objects, and you see there is really good structured data that specifies things like titles, actors, directors, genres — really good data that can make up the knowledge graph for this intelligent assistant.
Now you can let these crawlers run for hours or potentially days, and when you come back you can see that you'll have a knowledge graph that contains millions of unique nodes of really good structured metadata.
The next step you need to do to voice-enable your app or website is provide a front-end to your users to ask a question using the microphone on their devices and get an answer in return.