Train a Deep-Learning Text Classifier with Zapier



Watching this Video will take Approximately: 


Reading this lesson will take approximately: 


The Original Source of this content may include additional useful information or context, and will likely contain links to any content, templates or resources mentioned by the narrator. You can find visit the source below:

Lesson Content: has always looked at ways to help people with no technical or data science background to adopt AI and specifically Natural Language Processing.

Why Natural Language Processing, you may ask?

NLP is one domain of AI which can have a direct impact on your customer’s experience with your business which may ultimately result in higher retention.

Sounds good! What is the problem then?

The harsh truth is that getting these AI models to work requires substantial knowledge of coding, machine learning, and deep learning. And even if you have the prerequisite knowledge, it can still be a very daunting task. Extremely daunting. And cumbersome.

Introducing, Aasaanai for Zapier, a Zapier integration which enables a user to train and use state of the art deep learning text classifiers within 5 minutes! Without writing a single line of code.

How can the Aasaanai Zapier Integration help you? can help with faster Customer Support and Targeted Acquisition.

How? — We will be training a tweets sentiment classifier, from within Zapier, and use this model with other models in the Aasaanai marketplace to develop automation. This automation will enable our customer support team to be swift with their actions.

Let's start

Visit and press the create model button.

This will redirect you to the login page. Click the signup link and create an account.

After sign up, you will be redirected to the home page. You can either create models on the homepage itself or like we present later, within Zapier.

For using the Zapier Integration, we just need to grab the API key. Press the user icon on the top-right. Move to the API key tab and copy.

Train the Tweets Classifier

We are ready to train the classifier. Since, we are still in beta, you need to grab the Aasaanai Zapier link from here.

Follow these steps-

We will start by creating a new zap.

For the first application, we will use the Google Drive Trigger to upload the training data CSV file which is required to train the model.

Choose the trigger event as New File. (The idea is that the model is trained every time you add a new data CSV file to the Drive. So you need to ensure that the dataset is always the same for a particular model.)

Choose the file you want to use.

Now we will use the Upload CSV to Aasaanai action, to upload the data CSV.

Once we are done here, we can now use the Train text classifier action. This action takes in your email address to notify you when your model finishes training.

We now have a trained classifier!

Using the Tweets Classifier

We will be using the Tweets Classifier that we trained to keep track of tweets which are either negative or neutral so that the customer team can jump in to assist the affected users swiftly.

For this automation, follow these steps :

Start with the Search Mention in the Twitter trigger and select the twitter mention you want to use. For example, in this case we will be using the @instacart Twitter account.

We will then use the Language Detection Model from the Marketplace in order to ensure that the tweet language is in English.

We will then use the Tweets Classifier that we trained earlier to get the sentiment of the tweet.

We will now move forward if the sentiment is NEGATIVE.

If the tweet is negative, we will collect a few data points to deliver to the customer support team.

We will use the Summarizer Model from the Marketplace to get a summary of the tweet.

And the Entity Extraction Model to get the main entities from the tweet.

Routing to the relevant team

Finally, we will route the information to the Slack Channel for Internal Customer Support, hence enabling swift responses to negative tweets.

That's it!

How about targeted customer acquisition?

Follow the steps above, but instead of your twitter handle, use your competitor’s Twitter handle.

And when the model detects negative tweets, you can send automated tweets to the affected user, promoting your product and service!

We did this for @SwiggyCares and in addition to the above models, we also used a Keyphrases Extraction Model from the Marketplace to get the relevant phrases in the tweets.

The possibilities are endless. You can try to combine various NLP models suited for your needs and considerably improve and speed up your customer support pipelines.

If all this sounds interesting and exciting, you can become an early adopter (Early adopters receive special discounts, freebies and higher quotas for the free tier).

Tools used in this Tutorial