Parallel Pre-trained Model Pool for Document Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Cognitive capture systems face cold-start problems in document classification due to insufficient information for users or items, leading to inefficiencies in analyzing long documents without adequate annotations, and traditional methods like data crowdsourcing are costly and prone to mismatches.
Innovation Solution
A computer-implemented method that collects pre-trained classification models and documents into pools, applies these models in parallel to generate labels, and uses weighted voting to determine a final label result, enabling the creation of a baseline classification algorithm without additional data annotation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional data crowdsourcing methods are used for document classification, then classification capability can be achieved, but the cost increases and mismatches occur
Solution Approach 1:
The patent applies pre-trained classification models to documents before actual classification is needed. These models are trained in advance on general data and can be directly applied to new documents, eliminating the need for costly manual annotation while providing reliable classification results from the start.
Solution Approach 2:
The system uses automatically generated labels from pre-trained models to create baseline algorithms that can independently classify documents without human intervention. The generated labels serve the system itself, reducing dependency on external crowdsourcing and eliminating associated costs and mismatches.
2Productivity
If pre-trained models are applied in parallel to generate labels, then classification speed improves, but system complexity increases
Solution Approach 1:
The patent divides the classification task into multiple parallel processes, each handling a different pre-trained model. By segmenting the work across multiple independent model applications, the system achieves faster throughput while managing complexity through modular, independent components that can be executed concurrently.
Solution Approach 2:
The patent combines multiple pre-trained models and their generated labels into a unified baseline classification algorithm. By merging the results from parallel model applications through aggregation and voting mechanisms, the system achieves both speed improvements from parallelism and simplified output through consolidated classification decisions.
Data Source
AI summary
Label-based document classification using artificial intelligence includes collecting, by one or more processors, a plurality of pre-trained classification models into a model pool and a plurality of documents into a document pool. The collected plurality of pre-trained classification models are applied in parallel to the plurality of documents in the document pool to generate a list of labels. Based on the list of labels, a final label result is generated according to which a baseline algorithm for document classification is generated by the one or more processors.