AI Document Classification with User Intent Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI-based document classification methods struggle to accurately reflect the user's intention when classifying patent documents, despite their ability to understand the content well.
Innovation Solution
The proposed AI-based document classification method involves a computing device that reads documents, assigns user-selected documents to classification blocks, learns an AI model from these assigned documents, and classifies additional documents using this learned model, ensuring that the classification aligns with the user's intention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If AI technology is used to classify patent documents, then classification speed is improved, but classification accuracy reflecting user intention deteriorates
Solution Approach 1:
The system implements feedback by allowing users to review and correct AI-classified documents. User corrections are fed back into the system to retrain and improve the AI model, creating a continuous improvement loop that maintains both speed and accuracy over time.
Solution Approach 2:
The system performs preliminary classification using AI to quickly categorize documents, then allows users to review and adjust classifications before finalization. This preliminary action by AI followed by user verification resolves the contradiction by maintaining speed while ensuring accuracy.
2Measurement precision
If manual classification by patent attorneys is used, then classification accuracy reflecting user intention is improved, but classification speed deteriorates
Solution Approach 1:
The system merges AI-based automatic classification with human expert review into a hybrid approach. The AI handles the bulk of classification work for speed, while human attorneys review and correct classifications for accuracy, combining the advantages of both methods.
Solution Approach 2:
Instead of requiring full manual review of all documents, the system applies AI classification to all documents for speed, then applies partial human review only to documents that need correction or are particularly important, achieving both speed and accuracy efficiently.
3Adaptability or versatility
If AI model learns from all documents, then classification coverage is improved, but processing time and computational resources deteriorate
Solution Approach 1:
The system segments the document set into training documents and classification documents. The AI model learns from a segmented training subset to build classification capabilities, then applies this learned knowledge to the remaining documents for classification, reducing processing time while maintaining coverage.
Solution Approach 2:
The system performs preliminary learning from a subset of documents before actual classification begins. This preliminary action allows the AI model to develop classification skills in advance, enabling faster processing of the full document set while maintaining comprehensive coverage.
Data Source
AI summary
An AI-based document classification method and computing device are disclosed. The AI-based classification method comprises displaying an indication to represent whether an AI model can classify for the classification block selected by the user, and classifying documents selected by the user using the AI model, when a request for document classification with respect to the classification block is received from the user.