⚙️How it works

From our point of view, the best way to achieve our objectives is to build a web3 platform federating a community of users incentivized to produce good quality data for companies in need. Ta-da is the bridge allowing demand to meet supply in the data world. This section describes with the help of an illustrated example the general workflow proposed by Ta-da. In this fictional example, we consider that a company needs voice recording (basically thousand of users recording their speech when they read sentences) to train its voice assistant (like Google Home or Amazon Alexa). Other examples with other types of data could have been used: image collection for facial recognition, text retrieval for sentiment analysis, etc. Ta-da will handle each of them in the same way. Below is an overview of the workflow proposed by Ta-da:

Here are some details about each steps:

  1. A company needs English voice recordings to train the new voice assistant it is developing. It can use Ta-da to collect a great amount of good quality data. So it submits a job on the platform containing all its specific needs and criteria as well as a budget to pay users.

  2. Ta-da processes the job, splits it into many micro-tasks and sends them to users (the producers) who meet the criteria (English speaking users). In our example, a micro-task could be a simple English sentence to read. Thanks to the application, the user records himself by reading the sentence.

  3. Once the user is satisfied by his recording, he submits it. The data is directly sent to Ta-da. At this stage, we still don't know if the data is valid. A malicious user may have recorded a bad sentence or even nothing at all.

  4. To validate the data, it is sent to many other users from the community, who act as checkers. The data is sent along with a voting form that contains several questions such as whether there is any background noise and which sentence is being read. Each user listens to the recording and answers the questions in the form.

  5. When the checker is sure of his answers, he submits them to Ta-da. Thanks to the checkers' votes, Ta-da computes the outcome of the votes which allows to validate (or not) the data. Many technical details on this calculation are given in the Consensus calculation.

  6. In our example, the producer data is validated by the checkers. It is then returned to the company and all users who participated are paid in TADA tokens, thanks to the budget set by the customer.

This example is deliberately simplified and takes the most favorable case where the data is valid. However, several questions may have arisen in your mind: what if the data is invalid, who gets paid, what prevents users from producing bad quality data, etc. The following page answers these questions.

Last updated