Release Notes - 31 August 2022

In this release, we’ve redesigned our Workflow experience to give you superpowers - making it easier than ever to perform AI powered transformations on your unstructured data.

We've also added two new workflows to supercharge your tag editing experience. And improved Workflow performance by 60x! 🚅 🪄🪄🪄

But that’s not all. We’ve also given you more power over date filters, visualisations and tooltips.

Let's take a look inside! 👇

Your New Superpower: Workflows 🧙

What are Workflows?

Workflows are Relevance AI’s no-code tools for transforming unstructured data. In a few simple steps, you’ll be able to do things like AI tagging, sentiment analysis and category generation. For this release, we’ve given our workflow hub some much needed love!

Why does it matter?

Running workflows is a crucial step in analysing unstructured data in Relevance AI, unlocking most of our core features. Our product team has been focused on making them as easy to discover and use as possible.

How does it work?

When viewing a dataset in Relevance AI, look for the Workflows section in the left navigation panel.

You’ll be greeted by our Workflows hub, where you can browse through or search for workflows that solve your unstructured data needs.

Running a workflow is easy! Our workflow wizard will guide you through it step by step, so all you have to do is make simple choices.

Once you've finished following the workflow wizard, you have two options:

  1. Run on the Cloud
  2. Run in Colab

Running on the Cloud means that the workflow will run on Relevance AI servers. It will take longer to complete, but allow you to close the tab or do something else in the meantime (e.g. check emails, make a sandwich). You’ll be able to track the status of the workflow, but not the exact progress.

Running in Colab is the most stable and efficient option. It will open a new tab within a Google Colab notebook. All you have to do is copy and paste a token provided to you, and click the play button. You’ll have to keep your tab open while the Google Colab notebook completes, but there will be a progress bar.

When the workflow finishes, your AI transformation will be ready!

Workflow Performance: 60x speed-up! 🚅🚅🚅

The core improvement to workflows this week is our speed-ups. Over the next few days, we will be unrolling incredible speed-ups (in some cases up to 60X the speed on our largest datasets and largest models) to our workflows while ensuring that accuracy of our results remain the same.

Brand New Workflows!

New workflows in Relevance AI this week are supercharge tag editing:

  • Merging Tags
  • Deleting Tags

Improvements to existing workflows to attain better results:

  • Keyphrase generation
  • Sentiment

Other features in this release 💯

Quick date filters

It’s now easier to filter by date. If your data includes a date field and is current (e.g. last 2 years), you’ll have these options to quickly filter by:

Improved control over analyzing category metrics

A subtle but important feature!

You’re now able to visualise a different metric in your Category Overview than what you are sorting by. Note how in the screenshot below, we are sorting by the category label (alphabetically), but charting frequency:

Add tooltip annotations for your data fields

Want to provide your team mates more context about what a data field is?

You can now add field tooltip descriptions in your dataset Settings page.

Below I am adding an annotation that this “Primary Categories” field is "user generated tags".

Instead of having to guess or look up a data dictionary for defintions, users will see this when they hover over the field name in tables or documents.

Tip of the week ✨

Did you know that adding page specific filters will add a toolbar above your Explorer pages, with the filters easily accessible?

Set up a page specific filter in your left filter sidebar:

Now you will see a top filter bar above your Explorer pages for quick access:

Note that this only appears for page specific filters, not global filters.


Did this page help you?