Here are the main types of AI tasks you may encounter on our platform. Please bear in mind that we're continuously expanding the range of tasks that our workbench supports to perform a wider variety of projects. You may also receive email invitations from our Projects Team to tasks that are not described here yet!

While the list below will give you a general overview of the types of AI tasks available on our workbench, actual projects may have slightly different requirements. Always be sure to read the relevant Project Guidelines to understand the specific requirements of your task before starting.

 

Sentiment analysis (audio/text)

audio-sentiment-analysis.png

sentiment-analysis.png

Your role in this type of task is to read (or listen to) sentences or short paragraphs and categorize whether the author is talking about the topic in a positive, neutral, or negative way. Additionally, you may be asked to highlight instances of subtle nuances of language, such as sarcasm or irony. You may also be required to indicate how confident you are in your answer by rating your confidence as high, medium, or low.

Keep in mind you'll need to identify the overall attitude/tone the author is expressing (not your own feelings about the content).

 

Text/data categorization

data-categorization.png

image-categorisation.png

Text or data categorization involves reading a piece of text (or looking at an image, or listening to audio data) and marking the categories which apply from the provided options. The client will provide a classification system that specifies the level of detail they need. In some cases, multiple categories may apply.

One example here includes reading small excerpts of news articles and classifying them as “political”, “financial”, “humor” and others. 

 

Entity annotation

entity-annotation2.png

Entity annotation is the process of labeling different parts of a sentence (also called annotating entities) to enhance the reader's understanding of the text. These tasks involve reading a piece of text and adding labels to words or phrases within the text according to a specific class system. Some label examples are: countries, cities, or names of people. For example, you may need to differentiate the “apple” fruit from the “Apple” company, labeling one as “company” while the other as “food”. 

These types of tasks help AI systems better understand and parse text for the end goal of, for example, being able to automatically categorize whether a piece of news is about software or farming.

 

Image annotation (bounding boxes / polygons)

image-annotation.png

image-annotation-polygons2.png

Annotating images involves looking at the images provided and drawing shapes around certain parts of them, such as people, cars, or signs. If the shapes are  simple rectangles or squares, they are called bounding boxes, and if the task is to draw complex shapes by placing and connecting points around an image, these are called polygons. More detailed annotations might involve noting what is contained within the bounding box, particularly if the contents are text-based.

 

Keypoint annotation

keypoint-annotation.png

Also image-based, keypoint annotation requires you to mark certain points on the image, based on specific instructions provided for each type of project. One type of project in this category is marking where the various body parts are on a person as they move around. Think of marking the placement of right and left elbows and shoulders in a football player as they run towards the goal. 

 

Transcription / Video captioning

transcription-audio2.png

video-captioning.png

This task involves listening to a piece of audio or watching a video and typing out in text format what you hear. 

 

Have more questions? Submit a request

0 Comments

Article is closed for comments.
Powered by Zendesk