Watch videos with subtitles in your language, upload your videos, create your own subtitles! Click here to learn more on "how to Dotsub"


0 (0 Likes / 0 Dislikes)
[AI Business School] >> Pablo: This is a picture from an actual customer. As we engage with customers across a variety of verticals in the industries, what we find is that invariably we accumulate a huge amount of data, both structured and unstructured. And all the value that is latent in that data is lost in the complexities of the volume and the lack of structure in it. So in cases like this, we engage with these customers and we've been learning what is the best way to go from this to an application where you can ask questions and have structured data that has value to power your business. Of course it's not always the paper version of it. We see these very often in their digital form, where we accumulate years and years of unstructured content under structured counterparts. And they remain locked until we use patterns and lately, AI-powered patterns to extract the value out of them. And we call this activity knowledge mining. Here at Microsoft, we do it with a technology called cognitive search. In cognitive search,m we take this through a three-step process. First, we need to find all of your data and ingest it into a system. We'll do this on any type of data you may have in any data search in any format. Just bring the data to us; we'll deal with the details. Second, we need to enrich this data. We will use the variety of models we have in cognitive services, which is Microsoft pre-built AI solutions to crack, look and understand all of the details of your data. This could be for structured and unstructured content, it could be for pictures, it could be for documents. We have a variety of components that allow us to look at the content, no matter what the nature of the content is, and get the latent information that is inside. Once we used these AI components to understand what's in there, we produce annotations that are a former representation of the knowledge that we gathered that allows us to then enable a variety of application scenarios. We call that exploration. You can explore the data with different patterns from searching to doing analytics and more. When it comes to how do you actually bring AI to this picture, in order to understand the content, no matter what the nature of the content is, we want to make sure you can hit the ground running. So we have a broad set of built-in, ready-to-go cognitive skills powered by cognitive services that allows you to get going immediately. We'll do automatically things like key phrase extraction or entity recognition. We'll do face recognition, as well, everything from landmarks to celebrities. We can understand and describe an image, and much more. So all of these things allow you to get your content and take it into a place where you can project it into your application as structured data right away. Now, for the cases where you have an industry-specific need or something that is not built into our platform, we also support custom skills. These are usually built with one of our AI offerings, either Azure machine learning or data bricks or any of the alternatives, where you can bring your own custom AI component. But you don't have to now start from scratch. You can still fit it in the cognitive search scenario so that you don't need take care of any of the other details that are required to construct the end-to-end solution for knowledge mining. Now let's drill into the different exploration scenarios that you can use to consume the information we produced during the augmentation phase. First, searching interactive discovery. In this case, we use the technology that we already have on Azure, called Azure Search, that enables us to take all this informatiion that we discovered, including all the information from AI components, and put them in a search index where we can answer questions fast, no matter what volume of data we started with. This enables an interactive experience where our customers can ask questions and get answers quickly. Another way to consume this information is through knowledge graphs. In the end, when we look at this huge volume of information and you apply AI components to it, what we're trying to do is isolate the interesting bits of information. Think about entities, for example people and places and dates and organizations. Not only are we extracting those from the documents, but we also are connecting them. So now you can navigate an entire knowledge graph that helps you understand how everything is interconnected. And beyond that, once these things are in place, you can go further and overlay multiple data sources in a single larger graph that allows you to ask questions that go beyond any single data set or any single application. Finally, with enough volume of data together, you may also want to start asking analytical questions. When you have a large volume, aggregate-type questions are very typical and all of our tooling, powered by applications such as Power VI, comes into place to enable all of our analysts to answer interesting questions about business trends, trends of lower data, or what's happening out there that our data is representing. We are seeing customers in a broad set of industries apply this knowledge mining pattern to solve particular challenges they face. For example, in the legal domain, we see customers facing the challenge of having a huge amount of content to go over that involves extracting entities, people and places and times, and needing to connect all these dots to understand both legislation in a particular case and have answers quickly without having to wait or use a lot of bandwidth from paralegals. In the finance space, we have multiple customers that face the challenge of regulatory requirements for going over a huge number of forms and having to understand particular patterns in them. In the energy space, we've worked with multiple customers where their subsurface teams made of geologists need to go to a field and collect samples and take pictures, and they have years and years of that saved. And through knowledge mining, we can enable them to extract information from there and be able to reuse what they have in future efforts. Then finally, in the media space, we have engagements with sports teams, for example, where they want to track how they're coming across to their fans, what is happening out there on the media and in social media, and be able to connect individuals that they care about, fans and activities all in a single view, where all this unstructured data made of pictures and comments can be brought together into a single knowledge graph where they can ask questions and see how they're coming across to their audience. These are just a few examples of how customers are using knowledge mining to unlock all the latent value that they have in all their data.

Video Details

Duration: 7 minutes and 32 seconds
Language: English
License: Dotsub - Standard License
Genre: None
Views: 7
Posted by: csintl on Jun 18, 2019


Caption and Translate

    Sign In/Register for Dotsub above to caption this video.