Smart ETL with Data Prep: Detect Sentiment in Three Steps


The next-generation data platform is now generally available in Winter ‘21, and that means you get to embed sentiment analysis directly within your Data Prep Recipes!

What is Sentiment Analysis in Data Prep?

Data Prep introduces a new data transformation step, “Detect Sentiment”, that uses Natural Language Processing (NLP) and Machine Learning to extract sentiments in the form of “Positive”, “Neutral”, and “Negative” from your long text fields. To extract sentiment from your long text fields, you simply apply the “Detect Sentiment” transform on the desired fields in your recipe definition, and after the recipe job runs, it will add the extracted sentiment columns in your target dataset.

Why does my company need it?

If you don’t currently read customer survey data, customer case comments, or other forms of feedback that customers are providing to you via your websites, your channels, or social media, you’re likely missing out on opportunities to improve your products, services, or customer experiences. But perhaps you don’t have the time/budget to invest in an ML project for sentiment analysis due to your enterprise’s ever-growing competing priorities. Detecting Sentiment in Recipe is the perfect way to get started since it is so simple to build in recipes, it requires no new licenses or training for your teams. In addition, automating sentiment analysis in Data Prep as part of your ETL process means your data processing scales directly. Detecting sentiments on net new data means simply adding those data sources into a recipe, and you can start extracting sentiments from customer survey feedback, case interactions, CSAT responses, and whatever you can throw at Data Prep recipes, all the way to 2 billion rows of data.

That means you don’t need a new ML project or to kickstart a skunkworks ML team, you don’t need to budget for an ML system, infrastructure, project team, or system operators, and you don’t need a separate data pipeline to push-and-pull data between your source data, your ML system, and your analytical data platform in order to detect sentiments from your unstructured data for:

  • Customer Service / CSAT
  • Sales performance trend-identification and coaching
  • Social media brand management

OK, you convinced me, how do I use it in Data Prep?

1 – Create your data prep recipe with your input dataset.

Screen Shot 2020-10-05 at 8.40.40 PM.png

2 – Add a transform node, Select your long text field, click on “Detect Sentiment” transform, and Apply

Here they are step-by-step:

Add the transform step.

Screen Shot 2020-10-05 at 8.40.46 PM.png

Select your long text field.

Screen Shot 2020-10-05 at 8.43.32 PM.png

Click on the “Detect Sentiment” icon in the toolbar.

Screen Shot 2020-10-05 at 8.43.46 PM.png

Click on “Apply” in the left panel.

Screen Shot 2020-10-05 at 8.44.05 PM.png

That’s it!

3 – Save and Run your recipe

Screen Shot 2020-10-05 at 8.45.17 PM.png

4 – Analyze sentiments by customer segments, region, product types, channels, or however you please. You have the power!

Easy as Pie.

Give it a shot in your Winter ‘21 org and see how easy it is to incorporate sentiment analysis in your large volume ETL in Data Prep!

How useful was this post?

Click on a star to rate useful the post is!

Written by

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.