From the course: Microsoft Power Apps: AI Builder

Build and train the classification model - PowerApps Tutorial

From the course: Microsoft Power Apps: AI Builder

Build and train the classification model

- [Narrator] In our last episode, we added a survey comments entity to the comment data service in this environment by using Power Query to import this data from Excel. Now we're ready to build our model. Go under AI builder, click build. Remember that there are two sections of models here. And while they're not labeled this way on the page, they are indeed customized models at the top, refine a model for your business needs, and built in models at the bottom, get straight to productivity. This category classification model is ready to be used, but it wasn't trained with our data or our tags. We're going to use this customizable model here at the top. So click category classification. Name the AI model. And I'm going to call this Analyze Survey Comments. What you'll need. Text entries already classify with tags. 10 or more of each. Let's click create. The model is being set up. And next we'll be asked, where's our training data? Find your tagged text in the Common Data Service. Select text. So it's asking where is our text? Well, we know that it's under survey. There we go, using search. Survey comments. And we're being asked not about our tags here, but our text. Our text is in comment text field. Even though this says tagged text, to be clear, we're being asked for one field. And we're going to select this as having our text. And here's a preview so that we can make sure we're talking about the same data. Yep. That's our data. These were our customer comments. I'm going to click next, and next we'll be asked to identify the field that has the corresponding tags for this text. Point us to where you store your tags. Select tags. It will only go back to the same entity. Remember, we have to have both our tags and our text living in the same entity in the CDS. Not survey ID, tags. Select field. And again, we'll get a preview. We have automatically detected your separator and selected it for you. The comma, that's correct. Pay attention here. There are four possible tag separators. I told you there were three, comma, tab, and semicolon. The fourth is to have no separator at all. But in that case, each and every entry becomes its own tag. So rather than, for example, staff and safety being separated into two different categories, or two different tags, they're considered one, which is different than either staff or safety. So make sure that the proper separator has been selected, most of the time, it will be. And click next. I'm going to review the text and tags. Note some have one tag, some have two. And this is not all of our training examples, it's just a few to make sure that they look right. Because if they don't, we go back, and we make different choices. Click next. And one more choice. You choose the language that was used in our text column. All of our input text. So if we were to have surveys that had been answered in multiple languages, its best for us to set up different models. One per language. Click next. Make sure that this is what we want to do, we've made the right selections, click train. And I will see you after our models have been trained.

Contents