AI Data Collection Workflow With Context-Guided Labeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing artificial intelligence (AI) solutions require domain expertise for data collection, and non-expert users face challenges in collecting and integrating relevant data due to a lack of guidelines, leading to inefficient and resource-intensive manual processes.
Innovation Solution
A user-driven workflow engine guides non-expert users through a data collection procedure using a graphical user interface to efficiently collect and label data, incorporating context information and visual structures like bounding boxes, enabling precise AI applications with reduced time and resource consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing AI solutions are used, then meaningful results can be produced, but domain knowledge and expert skills are required for data collection and integration
Solution Approach 1:
The patent introduces an intermediary system comprising a workflow engine and contextual information manager that mediates between the user and the AI data collection process. This intermediary automatically generates contextual information, manages data collection workflows, and handles label generation, thereby eliminating the need for users to possess domain expertise while ensuring high-quality AI results through structured and automated processes.
2Adaptability or versatility
If manual data collection is performed without guidelines, then flexibility is maintained, but time and human resources are excessively consumed
Solution Approach 1:
The patent applies preliminary action by pre-defining workflow templates and contextual information structures before data collection begins. The system pre-generates relevant contextual information based on the selected workflow, and pre-establishes the framework for automatic label generation. This preliminary setup enables flexible adaptation to different data collection scenarios while significantly reducing the time and resources needed during actual data collection execution.
3Productivity
If data is collected without proper labeling, then collection speed increases, but the data cannot be effectively used by AI methods
Solution Approach 1:
The patent implements feedback mechanisms where the workflow engine continuously monitors the data collection process and automatically generates contextual information and labels based on predefined workflows. The system provides real-time feedback on data quality and completeness, automatically adjusting the labeling process to ensure AI-usability. This feedback loop enables high-speed data collection while maintaining reliability through automated quality assurance and contextual enrichment.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method of performing a data collection procedure for generating a dataset for input to a process which uses artificial intelligence comprises running, by a computing device, a workflow engine which is configured to perform a data collection workflow, wherein the workflow engine, when executing the data collection workflow, requests user input and guides the user through a user-driven data collection procedure of collecting labeled data and creates a dataset of labeled data as a basis for the process which uses artificial intelligence, defining a context of the data collection procedure and generating corresponding context information, after defining the context of the data collection procedure, executing the data collection workflow and providing, by the workflow engine when executing the data collection workflow, information relevant to the data to be collected on a graphical user interface, and using the context information, the provided information relevant to the data to be collected, and the received user input during the data collection workflow by the workflow engine for labelling the data and generating the dataset.