NLP Studio

Overview

Dataloop’s NLP (Natural Language Processing) Studio or Text Classification Studio offers a set of powerful text annotation tools that are essential when working with text data, enabling annotators to quickly produce a variety of meaningful annotations.

Before You Begin

Supported formats

Refer to the Supported file formats section to view the supported formats in the NLP Studio.

Task Prerequisites

Labeling work is performed in the context of a labeling task. Create your task with the intended data (image files), the Recipe (labels, attributes, and labeling instructions), and annotate team members.

Supported Annotation Tools

• Text Tool
• Text Classification

Refer to the Annotation Tools section to view the supported annotation tools formats in the Image Studio.

Annotation Task Configuration

The NLP studio configuration is controlled by the settings in the recipe.

1. Select recipes from the left-side panel in the Data Management section.
2. Locate the recipe and click it from the list.
3. Select the Instructions tab.
4. Locate the NLP studio-related items, such as the paragraph delimiter, and use pre-wrap spaces in text files from the Global Settings section.

Creating Annotations

1. In the dataset browser or task browser, double-click on the text file to open the file for annotation – the NLP studio will open the text item.
2. Select the Text tool (by default) from the annotation tools list. If you have only one annotation tool (defined in the recipe), it will be automatically selected.
3. Select text and clicking on the appropriate label from the labels. The selected text part is highlighted with the selected label.
4. When hovering over annotated text, you have the following options:
   1. Setting the annotation object ID.
   2. Cloning
   3. Deleting the annotation.
   4. Clicking the hamburger icon for Fast classification. Fast classification will apply the selected annotation to all identical words in the text.

Text Classification

The Classification tool enables applying a label class to an entire text document.

1. Click on the classification tool.
2. Click the label(s) from the label list to create a classification annotation of the text. The newly created annotation can be seen in the annotation list on the right side of the screen and as a label above the text.

Auto Next Item

The Auto next item checkbox appears when the Classification tool is selected.

When the task requires only a single label per text, checking the Auto next item feature will automatically save the annotation and load the next item as soon as a label is selected.

Paragraph Delimiter

Setting a paragraph delimiter allows you to choose whether you wish to annotate individual words or sections of the text set by the delimiter. The paragraph delimiter can be any character, including the space bar or Enter key.

By default, the paragraph delimiter is empty.

To set the paragraph delimiter, follow these steps:

1. Under the Instructions tab, click the field next to the paragraph delimiter.
2. Set your delimiter of choice. For example, type a character or press Enter or the space bar to set the paragraph delimiter.
3. Save the changes you made to the recipe by clicking the Save icon.

Once a paragraph delimiter is set, each annotation you make will apply to the entire delimited section.

To clear the delimiter setting, click the X button next to the paragraph delimiter field.

Note: When the paragraph delimiter is set to None, the text will be presented as one paragraph. If you wish to have the text in its original paragraph structure, check the box “use pre-wrap spaces in text files.”

Parent-Child Relationship

This useful annotation tool allows the creation of relationships between annotations.

For example, in the sentence “David used a hammer, some nails, and a saw to build this beautiful table,” if you wish to show that the tools belong to David, you can set David as the parent and the tools as the children.

To create a parent-child relationship between annotations, simply click and hold on to an annotation and drag it to its parent.

This relationship can be seen in the annotation list when clicking on the Annotation Grouping icon and when hovering over the annotation with the parent-child relationship.

Another way of setting a parent-child relationship between words is by selecting an annotation or group of annotations from the Annotations tab on the right.

Once an annotation or annotation group is selected, hover over the annotation and then over the text that appears above it.

Click the Set as Parent icon to set the annotation as a parent of the annotations selected from the annotation list.

Once an annotation is set as a parent, the option to disconnect all children also appears when hovering over the annotation.

Learn more about working with the Studio 2.0 annotation list.

Action Icons

• The Annotation Focus icon filters the annotation list to include only annotations currently visible on the screen.
• The Annotation Grouping icon toggles between listing the annotations one by one or by label group.
• The Annotation Sort icon allows sorting the annotation list by creation time, update date and time, label, or type.

Hover over any annotation to see related actions:

• Change the annotation label.
• Toggle visibility using the eye icon.
• Open issues on an annotation.
• Move an annotation for review.

You can read more about how these icons are used in quality assurance (QA) assignments here.

NLP Studio Keyboard Shortcuts