AI Data Labeling via Iterative Clustering Refinement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data labeling methods for clustering tasks are inefficient and inaccurate due to the high workload of manual labeling, especially when dealing with large datasets, which limits the accuracy of clustering labeling.
Innovation Solution
A data labeling method based on artificial intelligence that iteratively pre-clusters samples and refines class clusters using labeling information, reducing the workload by maintaining smaller sample sizes within each class cluster, thereby improving labeling accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual labeling is used for large datasets, then clustering accuracy can be improved, but the workload and time consumption increase significantly
Solution Approach 1:
The patent segments the large dataset into multiple smaller batches or subsets that can be processed independently. By dividing the overall labeling task into manageable segments, the system reduces the time and computational resources required for each labeling operation while maintaining overall clustering accuracy through iterative refinement across segments.
Solution Approach 2:
The patent performs preliminary clustering operations on the dataset before final labeling. This preliminary action groups similar samples together, reducing the complexity of subsequent labeling tasks. By pre-organizing the data structure, the system minimizes the time required for manual labeling while preserving the ability to achieve accurate clustering results.
2Manufacturing precision
If manual labeling is performed on large quantities of samples, then clustering results can be improved, but the workload becomes too large
Solution Approach 1:
The patent implements self-service mechanisms where the clustering system automatically performs preliminary grouping and organization of samples based on existing features. This automated self-service reduces the manual workload required for labeling while maintaining high accuracy, as the system handles the tedious segmentation and preliminary classification tasks without human intervention.
Solution Approach 2:
The patent replaces manual mechanical labeling processes with automated computational clustering algorithms. By substituting human manual work with algorithmic processing, the system dramatically increases labeling efficiency and productivity while maintaining or improving accuracy through consistent, repeatable clustering operations that eliminate human error and fatigue.
3Reliability
If all samples are clustered together in one batch, then comprehensive clustering can be achieved, but the complexity and computational burden increase
Solution Approach 1:
The patent divides the comprehensive clustering task into multiple smaller, manageable clustering operations performed on segmented batches of samples. This segmentation reduces the computational complexity of each individual clustering operation while maintaining comprehensive coverage of the entire dataset through iterative processing of all segments, thereby simplifying the overall process without sacrificing completeness.
Solution Approach 2:
The patent employs dynamic clustering strategies where the clustering parameters, batch sizes, and processing methods are adjusted based on the characteristics of the data being processed. This dynamic approach allows the system to optimize computational resources for each specific clustering task, reducing overall complexity while ensuring reliable and comprehensive clustering results across diverse data types and scales.
Data Source
AI summary
Provided is a data labeling method based on artificial intelligence, an apparatus, and a storage medium relating to the field of artificial intelligence, particularly data labeling, image recognition, and natural language processing. The method includes: determining a plurality of samples involved in clustering; performing a plurality of following operations circularly to realize iterative processing, until a convergence condition is satisfied or a quantity of iterations reaches a number threshold, comprising: pre-clustering the plurality of samples according to a vector representation of the respective samples to obtain a plurality of class clusters, each class cluster containing at least one sample; receiving labeling information for the respective class clusters and re-determining the plurality of samples according to the labeling information; and determining a clustering result according to the labeling information for the respective class clusters.