Azure’s Cognitive Services are very easy to use from within your own applications or more “code-less” solutions such as Azure Logic Apps. In this post, I will show a simple example of a Logic App that does sentiment analysis on incoming tweets. When the sentiment score is very high, an SMS is sent.
To perform sentiment analysis, use the Text Analytics API from Cognitive Services. First, in the Azure Portal, create a Cognitive Services account of type Text Analytics. After that account has been created, you will need the following information to be used in Logic Apps:
- Endpoint: https://westus.api.cognitive.microsoft.com/text/analytics/v2.0
- Key: use one of the two secret keys to access the API
If you create a Cognitive Services account in the free tier, you can make 5000 calls per 30 days.
Now create a Logic App in the Azure Portal and go to Triggers and Actions to enter the designer. I will not provide step-by-step details as working with triggers and actions in the graphical designer is very easy. To obtain the results we want, we will have to switch to Code View though.
First, from the Microsoft Managed APIs, add a Twitter trigger. Provide your credentials to Twitter and provide a search term.
Now click the + icon to add an action. To call the Sentiment Analysis API, use the HTTP action and provide the following information (Note: forgetting to specify the specific API to call is a common error; for the URI below also add /sentiment to perform sentiment analysis!!!):
Naturally, replace your key with one of the keys obtained from your Cognitive Services account. The body of your HTTP POST can be an array of documents, with each document having an id and the text you want to analyze. In our case, we want to analyze the Tweet text so we use the graphical designer to insert it. In Code View, this will be:
Now we want to send an SMS when the sentiment of the Tweet is very positive. The sentiment is expressed as a value between 0 and 1 where 1 is a really, really, really positive tweet.
To send an SMS when the sentiment is above 0.95, first click the + icon and add a condition. The value to evaluate is part of the HTTP body of the previous action. So add that and select greater than or equals and enter 0.95 in the value. Then switch to advanced view to see the expression you built. It will look like:
@greaterOrEquals(body(‘Http’), 0.95)
The above is not going to cut it though since the response body is JSON and the sentiment score needs to be extracted from it. Change the expression to the following:
@greaterOrEquals(float(json(string(body(‘Http’))).documents[0].score), 0.95)
Since the response body is an array of documents and we only have one document, just obtain the score from the first document.
Now we can click Add an action in the If yes section to send an SMS. You can use the Twilio Send Message Managed API to do so but you will need an account at Twilio for this to work. Alternatively, you can send an e-mail or just post a result to http://requestb.in. For Twilio, you will end up with something like below. Phone numbers have been blurred to protect the innocent.
In the above, we only want to show the score for the Tweet and not the whole body. This can be done in Code View:
In the Send_Message action, change the following:
@{body(‘Http’)} for @{triggerBody()[‘TweetText’]}
to:
@{json(string(body(‘Http’))).documents[0].score} for @{triggerBody()[‘TweetText’]}
Note that changes like the above can make the UI designer unavailable.
When you save this Logic App, incoming tweets containing Azure should be analyzed and you should get SMSs when tweets are very positive. Hey, it’s Azure, why shouldn’t they be? 🙂
You can check if the Logic App is executing correctly from the Operations tile:
For a search term like Azure, I recommend to turn off the Logic App if you don’t want to exhaust your 5000 free tier API calls.
baeke.info 