Using the Cognitive Services Text Analytics API in PowerApps

This post shows you how to build a basic app that uses the Microsoft Cognitive Services Text Analytics API. We'll show you how to set up the Text Analytics API, and connect to it with the Text Analytics connector. Then we'll show you how to create an app that calls the API.

Introduction to Microsoft Cognitive Services

Microsoft Cognitive Services are a set of APIs, SDKs, and services available to make your applications more intelligent, engaging, and discoverable. These services enable you to easily add intelligent features – such as emotion and video detection; facial, speech and vision recognition; and speech and language understanding – into your applications.

We’ll focus on "language understanding" for the rest of this post, working with the Text Analytics API. This API enables you to detect sentiment, key phrases, topics, and language from your text. Let's get started by trying out a demo of the API, then signing up for a preview version.

Try out the Text Analytics API

The API has an online demo – you can see how it works, and look at the JSON that the service returns.

1. Go to the Text Analytics API page.

2. In the See it in action section, use the example text, or enter your own text. Then click Analyze.

3. The page shows formatted results on the Analyzed text tab, and the JSON response on the JSON tab.

Sign up for the Text Analytics API

The API is available as a free preview, and it is associated with an Azure subscription. You manage the API through the Azure portal.

1. If you don’t already have an Azure subscription, sign up for a free subscription.

2. Sign in to your Azure account.

3. Go to the Create Cognitive Services blade in the Azure portal.

4. Enter information for the Text Analytics API, like in the following image. Select the F0 (free) pricing tier.

5. In the lower-left corner, click Create.

6. On the Dashboard, click the API that you just created.

7. Click Keys.

8. Copy one of the keys on the right of the screen. We will use this key later when we create a connection to the API.

Build the app

Now that we have the Text Analytics API up and running, we can connect to it from PowerApps, and build an app that calls the API. This is a single screen app that provides functionality similar to the demo on the Text Analytics API page. Download the completed app, or get started on building it.

Create the app and add a connection

First, we'll create a blank phone app and add a connection with the Text Analytics connector. If you need more information about these tasks, see Create an app from scratch and Manage your connections in PowerApps.

1. In PowerApps Studio, click File > New, then under Blank app, click Phone layout.

2. In the middle pane, click connect to data.

3. In the right pane, click New connection > Text Analytics.

4. Copy your key into Account Key, then click Create.

Add controls

The next step in creating the app is to add all the controls. Normally when I build apps, I add formulas to the controls as I go, but in this case I'll focus on the controls first, then add a few formulas in the next section. The following image shows the app with all the controls.

Follow the steps below to create this screen. If a control name is specified, that name is used in a formula in the next section.

1. Add a Rectangle control (under Insert > Icons) as a background for the title, then add a Label control for the title. Add another Label control for the introductory text.

2. Add a Text input control, so we can enter text to analyze. Name the control tiTextToAnalyze. The app should now look like the following image.

3. Add three Check box controls, so we can choose which API operations to perform. Name the controls chkLanguage, chkPhrases, and chkSentiment.

4. Add a button, so we can call the API after selecting which operations to perform. The app should now look like the following image.

5. Add three Label controls. The first two will hold results from the language and sentiment API calls; the third is just an introduction for the gallery at the bottom of the screen.

6. Add a Blank vertical gallery control, then add a Label control to the gallery. The gallery will hold results from the key phrases API call. The app should now look like the following image.

We're keeping this app simple to focus on calling the Text Analytics API, but you could add things – like logic to show and hide controls based on the check boxes selected, error handling if the user doesn't select any options, and so on.

Add formulas

OK, we have a nice-looking app, but it doesn't do anything yet. We'll fix that now. But before we dive into the details, let's understand the pattern that the app follows:

1. The app makes specific API calls based on the checkboxes selected in the app. When we click Analyze text, the app makes 1, 2, or 3 API calls.

2. Data that the API returns is stored in three different collections: languageCollect, sentimentCollect, and phrasesCollect.

3. The Text property for two of the labels, and the Items property for the gallery, are updated based on what's in the three collections.

With that background, let’s add the formula for the OnSelect property of our button. This is where all the magic happens.

If(chkLanguage.Value=true,

ClearCollect(languageCollect, TextAnalytics.DetectLanguage({numberOfLanguagesToDetect:1, text:tiTextToAnalyze.Text}).detectedLanguages.name)

);

If(chkPhrases.Value=true,

ClearCollect(phrasesCollect, TextAnalytics.KeyPhrases({language:"en", text:tiTextToAnalyze.Text}).keyPhrases)

);

If(chkSentiment.Value=true,

ClearCollect(sentimentCollect, TextAnalytics.DetectSentiment({language:"en", text:tiTextToAnalyze.Text}).score)

)

There's a bit going on here, so let's break it down:

• The If statements are straightforward – if a specific check box is selected, make the API call for that operation.
  • Within each call, specify the appropriate parameters:
  • In all three calls, we specify tiTextToAnalyze.Text as the input text.
  • In DetectLanguage(), we hard-code numberOfLanguagesToDetect as 1, but we could pass this parameter based on some logic in the app.
  • In KeyPhrases() and DetectSentiment(), we hard-code language as "en", but we could pass this parameter based on some logic in the app. For example, we could detect the language first, then set this parameter based on what DetectLanguage() returns.
• For each call that is made, add the results to the appropriate collection:
  • For languageCollect, we add the name of the language that was identified in the text.
  • For phrasesCollect, we add the keyPhrases that were identified in the text.
  • For sentimentCollect, we add the sentiment score for the text, which is a value of 0-1, with 1 being 100% positive.

Now that we have the API calls, let's move on to displaying results. To display the results of the API calls, reference the appropriate collection in each control:

• Set the Text property of the language label to:

"The language detected is " & First(languageCollect).name

The First() function returns the first (and in this case only) record in languageCollect, and we display the name (the only field) associated with that record.

• Set the Text property of the sentiment label to:

"The sentiment score is " & Round(First(sentimentCollect.Value).Value, 3)*100 & "% positive"

This formula also uses the First() function, gets the Value (0-1) from the first and only record, then formats it as a percentage.

• Set the Items property of the key phrases gallery to:

phrasesCollect

We're now working with a gallery so we don’t need the First() function to extract a single value. We reference the collection, and the gallery displays the key phrases as a list.

Run the app

Now that the app is finished, let's run it to see how it works. In the following image, all three options are selected, and the text is the same as the default text on the Text Analytics API page.

You can compare this to the Text Analytics API page, and see that we get the same results.

Summary

We hope you now understand a little more about the Text Analytics API, and you've enjoyed seeing how to incorporate it into an app. Let us know if there are other Cognitive Services (or other services in general) that you would like us to focus on in our posts. As always, please leave feedback and any questions in the comments.