What are Data Annotations? The Ultimate Power BI Guide

Data, in its raw state, is not usable, and it is somewhat impossible to draw insights from it. To make it usable and valuable to the reader, you will have to incorporate additional elements like labels, metadata, or tags into it and that’s where data annotation comes in. But what are data annotations?

Data annotation is an expensive and complex process that should be handled by experts. There are lots of pieces that go into it, and lots of companies fail to put these pieces together.

In this guide, you’ll discover what are data annotations, data annotation tools, why data annotation is helpful, and other useful details about data annotation.

Table of Contents:

1. What are Data Annotations?
2. Why are Data Annotations Helpful?
3. How Many Types of Data Annotation?
4. What is the Difference Between Data Annotation and Data Labeling?
5. How to Choose the Right Tool?
6. How to Perform Data Annotations in Power BI?
7. How to Create Visualizations in Power BI?
8. Wrap Up

First…

What are Data Annotations?

Definition: Data annotation is the labeling of data to make it valuable for machine learning. It involves adding labels, metadata, or tags to raw data. Raw data could be video, audio, images, or text.

If you’re wondering, “What are data annotations,” they are the key elements that help machine learning models learn and understand the data. This will, in turn, boost their performance and accuracy. For example, during image recognition, annotators help in labeling the objects within the images. The labeling could be “car,” “dog,” or “cat.”

Data annotators are used in natural language processing to mark named entities or parts of speech in the text.

Why are Data Annotations Helpful?

Data annotations help in providing the labeled and structured data needed for training machine learning models. With data annotation, the models are taught to recognize trends and make accurate classifications or predictions that align with the patterns. For instance, trend analysis examples in data annotation include identifying seasonal patterns in sales data or detecting anomalies in financial transactions.

When considering the question, “What are data annotations,” it becomes evident that they are essential for ensuring the quality and accuracy of machine learning outputs. AI for data analytics relies on these annotations to enhance model performance. During image recognition, the labeled images help the model correctly categorize and identify objects. All these are necessary in applications like medical imaging, autonomous driving, and facial recognition.

Annotated text plays a role in natural language processing it enables models to understand language nuances, entities, and sentiments. This, in turn, helps improve tasks like sentiment analysis, chatbot interactions, and translation.

How Many Types of Data Annotation?

The common types of data annotation are:

Image Annotation

• Bounding Boxes: Involves identification of the location of objects by drawing rectangles around them.
• Semantic Segmentation: This involves the labeling of each pixel in an image with a class.
• Instance Segmentation: Involves differentiating objects within a class.
• Polygon Annotation: Involves the outline of complex shapes using polygons.

Text Annotation

• Named Entity Recognition (NER): This involves labeling and identifying entities like dates, locations, and names.
• Sentiment Annotation: Used in determining sentiment expressed in text.
• Intent Annotation: Used for classifying the purpose (or intent) behind a text.

Audio Annotation

• Transcription: Used for the conversion of speech to text.
• Speaker Identification: Used for labeling speakers in an audio file.
• Sound Event Detection: Used for identifying specific sounds (or events).

Video Annotation

• Frame-by-Frame Annotation: Used for labeling objects in each frame of the video.
• Object Tracking: Used to follow objects across different frames.

3D Point Cloud Annotation

• Object Detection: Used for the labeling of objects in 3D space and it can be applied in autonomous driving.

What is the Difference Between Data Annotation and Data Labeling?

When considering the question, “What are data annotations,” it is important to note that data annotation and data labeling are often used interchangeably but they’re different terms. Here are the differences between the two.

• Data annotation: It involves the identification and marking of various elements within a data. These elements could be part of speech in a text, sounds in audio, or objects in images.
• Data labeling: It involves assigning predefined categories (or labels to data points). For instance, labeling could be categorizing a piece of text as “not spam,” or “spam.”

How to Choose the Right Tool?

Here are things to consider when choosing the right tool.

• Type of Data

The tool has to support the data type you want to annotate.

• Ease of Use

There have to be support documents and tutorials. Also, the tool has to be user-friendly.

• Cost

Take a look at the pricing, and opt for a tool that fits your budget.

How to Perform Data Annotations in Power BI?

Here are easy ways to perform annotations in Power BI.

Using Text Boxes

• Add Your Text Box: Navigate to the “Insert” tab of the Power BI Service toolbar or Power BI Desktop.
• Insert Text Box: Click the “Text box.” A text box will be displayed on the report canvas.
• Enter the Annotation: Input your comment or annotation into the text box. Text formatting can be done using tool options like alignment, color, or font size.
• Position the Text Box: You can drag and resize the text box to your desired position.

Annotations on Visuals

• Title and Subtitle: To add descriptive titles and subtitles, choose your preferred visual, and navigate to the “Format” pane. Expand the “Title” section. Input the annotation into the Title or subtitle fields.
• Tooltips: When a user hovers over a data point, tooltips offer additional context. Customizing the tooltips involves selecting the visual and navigating to the “Format” pane. After that, expand the “Tooltip” section.
• Data Labels: Data labels are used for annotating individual data points. Choose your visual, navigate to the “Format” pane, and turn on “Data labels.”

Using Shapes

• Add a Shape: Navigate to the “Insert” tab and choose a shape. The shape could be a rectangle, arrow, line, or oval.
• Draw the Shape: Draw your shape on the report canvas to annotate or highlight specific areas.
• Add Text: To add text, select the shape and enter the text directly.

How to Create Visualizations in Power BI?

Here is a step-by-step process for creating a visualization in Power BI.

Stage 1: Logging in to Power BI

• Log in to Power BI.
• Enter your email address and click the “Submit” button.

• You are redirected to your Microsoft account.
• Enter your password and click “Sign in“.

• You can choose whether to stay signed in.

• Once done, the Power BI home screen will open.

Stage 2: Creating a Data Set and Selecting the Data Set to Use in Your Chart

• Go to the left-side menu and click the “Create” button.
• Select “Paste or manually enter data“.

• We’ll use the HR sample data below for this example.

Total Cost	Company Type	Company Name	Expertise Categories	Expertise	Cost
Total Cost	Subcontractor	Skyline Contractors	Mechanical Installation	Plumbing & Heating	15456
Total Cost	Subcontractor	Skyline Contractors	Mechanical Installation	Mechanical Work	10159
Total Cost	Subcontractor	Onyx General Contractors	Mechanical Installation	Plumbing & Heating	18045
Total Cost	Subcontractor	Onyx General Contractors	Mechanical Installation	Mechanical Work	12695
Total Cost	Subcontractor	Living Well Remodeling	Mechanical Installation	Plumbing & Heating	14589
Total Cost	Subcontractor	Living Well Remodeling	Mechanical Installation	Welding	11456
Total Cost	Supplier	Power-up Builders	Raw Material	Cement	20561
Total Cost	Supplier	Power-up Builders	Raw Material	Steel	32456
Total Cost	Supplier	Five-star Construction	Raw Material	Bricks	10253
Total Cost	Supplier	Five-star Construction	Raw Material	Timber	9000

• Paste the data table above into the “Power Query” window. Next, choose “Create a dataset only.”

• On the left-side menu, click “Data Hub.” Power BI will populate the data set list. An error message will be displayed if a data set has not been created.

• Select the data set to be used to create the Sankey diagram. Power BI will populate the screen as seen below.

• Click “Create a report ” from the dropdown menu” and choose “Start from scratch.”

• A Report Canvas screen will be displayed below.

Stage 3: Add Power BI Sankey Diagram Extension by ChartExpo

• Look out for the Power BI visualization panels, and open them. It’s found on the right side of your Power BI dashboard. Click the ellipsis symbol (…) to import the Power BI Sankey Diagram extension by ChartExpo.

• Click the “Get more visuals’ option.

• Input “Sankey Diagram for Power BI by ChartExpo” into the search box. You’ll see the “Sankey Diagram for Power BI by ChartExpo” displayed on your screen.

• Select the “Add’ button.

• The “Sankey Diagram for Power BI by ChartExpo” icon will be added to the visualization panel.

Stage 4: Draw a Sankey Diagram with ChartExpo’s Power BI extension

• From the visualization panel, click the “Sankey Diagram for Power BI by ChartExpo” icon. A window similar to the one below will be opened.

• There’s the option of resizing the visual. On the right side of the Power BI dashboard, there’s the “Fields” option next to the “Visualizations.”

• At this point, you’ll have to choose the fields in the Sankey chart. To do that, follow the sequence below:
• • Total Cost
  • Company Type
  • Company Name
  • Expertise Categories
  • Expertise

• You’ll have to provide your email address or ChartExpo license key.

Stage 5: Apply a Subscription Key or Activate the ChartExpo

• Select ChartExpo visual, and look out for the three icons below “Build Visual” in the Visualization panel.

• Choose the middle icon (”Format visual”) to get the visual properties populated.

• As a new user, you can start using ChartExpo by;
• • Input your email address in the textbox that’s found under “Trial Mode.” The License key will be sent to the email by ChartExpo. Your email address has to be up-to-date and accurate. Toggle the “Enable Trial” icon to activate the 7-day trial.

• You’ll get a welcome email from ChartExpo. After that, you can proceed to create a Sankey Diagram. But all charts created within the 7-day trial will have the ChartExpo watermark.

• If you have a license key, navigate to the “License Settings” section and input it into the “ChartExpo License Key” textbox. Look out for the toggle switch next to the “Enable License” and slide it “On.”

• The Sankey diagram should be ready. And charts created will not have a watermark.

• To add a Prefix (like the $ sign), you’ll have to expand the “Stats” properties and insert the Prefix value.

• To add colors to the node, you’ll have to expand the “Level Colors” properties and choose the colors.

• All changes will be saved automatically.

Insights

Here are some insights you can get from the chart.

• The procurement cost at Level 1 (Total Cost) is $155k.
• Out of the $155k cost, allocations were made at Level 2 (Company Type). These include $72.3k (46.7%) allocated to the supplier and $82.4k (53.3%) spent on subcontractors.
• At Level 3 (Company Name), the supplier cost ($72.3K) was divided between Five-star Construction and Power-up Builder. The charges were $19.3k and $53.0k, respectively.

FAQs

What is the annotation feature in Power BI mobile?

The annotation feature in Power BI mobile enables users to highlight, add notes, and draw directly into the Power BI dashboards and reports. It also enables collaboration with multiple users.

How do I add comments in Dax?

You can add comments in Dax ‘#’ using multi-line comments or double-forward slashes // for single-line comments.

What is the purpose of data annotation?

Data annotation highlights key points, clarifies insights, adds context, and improves the overall visualization experience.

Wrap Up

Data annotation involves the addition of extra elements to a raw dataset to enhance its readability and usability. It improves the accuracy of machine learning models and ascertains better decision-making in visualizations.

Data annotation and data labeling are used interchangeably, but they’re quite different. When asking “What are data annotations,” it’s important to note that data annotation involves making raw data more meaningful for machine learning, while data labeling involves assigning predefined categories or labels to data points.

You can improve annotations by:

• Using text boxes
• Adding tiles and subtitles
• Using tooltips
• Adding and drawing shapes
• Taking advantage of bookmarks and selections
• Using visual headers
• Adding comments.

The common types of data annotation include image annotation, audio annotation, text annotation, 3D point cloud annotation, and video annotation.

Now you know what data annotation is, what elements will you be incorporating into your data?