Language Data Labeling Using Topic Embeddings and Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Managing and organizing large numbers of language data structures stored on a computing system is challenging, as existing methods lack effective ways to label and present them efficiently.
Innovation Solution
A method and system utilizing a combination of a large language model and a clustering model to assign topics to language data structures, generate vector data structures, and cluster them based on semantic similarity, allowing for efficient organization and labeling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If manual organization methods are used for language data structures, then labeling precision can be maintained, but productivity decreases due to the large number of files requiring manual management
Solution Approach 1:
The patent introduces an intermediary system comprising a language model, encoding model, and clustering model that acts as a mediator between the large number of language data structures and the organization task. This intermediary automatically generates topics, converts them to vector representations, clusters similar topics, and applies cluster labels to datasets, thereby resolving the contradiction by providing automated processing that maintains organizational accuracy while dramatically improving productivity
Solution Approach 2:
The patent creates vector copies (embeddings) of the topics generated from language data structures. Instead of directly processing and labeling the original language data structures, the system creates vector representations that capture the semantic meaning, clusters these vectors, and then uses the cluster information to label the original data structures. This copying approach enables efficient automated processing while preserving the semantic accuracy needed for proper organization
2Productivity
If automated labeling methods are implemented, then productivity increases, but device complexity increases due to the need for multiple machine learning models
Solution Approach 1:
The patent segments the automated labeling task into three distinct functional modules: a language model for generating topics from datasets, an encoding model for converting topics to vector representations, and a clustering model for grouping similar vectors and generating labels. This segmentation allows each model to specialize in a specific function, making the overall complex system more manageable and maintainable while achieving high productivity through automation
Solution Approach 2:
The patent creates a multi-functional system where the combination of language model, encoding model, and clustering model serves multiple purposes: topic generation, semantic representation, similarity-based grouping, and automated labeling. This universal system can handle various types of language data structures and organization tasks, justifying the increased device complexity by providing versatile automated processing capabilities
Data Source
Figure 1
Figure 2
Figure 3A~3C
AI summary
A method including applying a language model to datasets to generate topics assigned to the datasets. Each of the topics includes at least one of a natural language text word and a natural language phrase. The method also includes applying an encoding model to the topics to generate a corresponding vector data structures storing embedded topics. Each embedded topic of the embedded topics is associated with one corresponding vector in the vector data structures. The method also includes applying a clustering model to the vector data structures to generate a cluster including a subset of the vector data structures. The subset includes a reduced number of the vector data structures. The method also includes modifying, according to the cluster, the datasets.