Data Sets with CuratorMAS will now be Edited by Automatic Curing Agents

The success of machine learning models depends directly on the quality of the data set used. However, the creation and editing of high-quality data sets is still expensive, time-consuming and hard to generalize domain from domain. To solve this problem, researchers developed a new framework called CuratorMAS.

Basic problems of existing solutions are:

• Manually designed heuristic methods can not be applied to different tasks and query types
• The approaches based on model-dependent signals fail outside certain scenarios
• Cost-effective value for large-scale projects that require human intervention Tags
• Systems trained in a domain cannot be moved to other areas
• Flexible adaptability to user requirements

CuratorMAS is a automation system designed to coordinate multiple AI agents. The system automatically evaluates, cleans and organizes data sets by enabling different specialized agents to move together. In this way, both costs are reduced and adaptable to different tasks.

The research points the beginning of the transition to the machines from the future of the data generation. Automatic systems such as CuratorMAS can enable data scientists focus on higher value jobs and reduce resource waste in model training.