Document Tagging Model Retraining Using Anonymized Feature Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current tagging models in document management systems inaccurately tag documents and are not adapted to user feedback, often using private information without permission, leading to potential data leaks.
Innovation Solution
Retrain machine-learned models using anonymized data by generating a feature vector representation of the target document, querying a corpus of publicly available documents, and creating a second training set of tagged documents to improve tagging accuracy without compromising privacy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If tagging models are trained with user documents to improve tagging accuracy, then model performance is improved, but user privacy is compromised due to use of private information without permission
Solution Approach 1:
The patent extracts only the necessary structural and semantic features from user documents while removing all personally identifiable information. Feature vectors are generated that capture document characteristics (layout, text patterns, formatting) without retaining private content, enabling model training that respects user privacy while maintaining tagging accuracy.
Solution Approach 2:
The patent introduces an intermediary processing layer that transforms raw user documents into anonymized feature representations. This intermediary step uses natural language processing and feature extraction techniques to create intermediate data structures that preserve informational content for training while eliminating direct access to private user information.
2Adaptability or versatility
If conventional tagging models are used to identify document components, then tagging function is provided, but tagging accuracy is insufficient and models cannot adapt to user feedback
Solution Approach 1:
The patent implements a feedback mechanism where user corrections and annotations of incorrectly tagged portions are captured and used to generate additional training data. The system continuously learns from user feedback by incorporating corrected tags into updated training sets, enabling the model to adapt and improve tagging accuracy over time while maintaining privacy through continued use of anonymized features.
3Reliability
If private information is used for model training, then model performance is improved, but data security is compromised and data leaks become vulnerable
Solution Approach 1:
The patent creates simplified copies of user documents in the form of feature vectors that retain only the essential structural and semantic information needed for training. These copies exclude all sensitive personal data while preserving document characteristics necessary for accurate tagging, thereby enabling model training without exposing private information to security risks.
Data Source
AI summary
A document management system trains a machine-learned model using a first training set of tagged documents to, when applied to a document, tag one or more portions of the document. The document management system applies the machine-learned model to a target document. One or more portions of the target document incorrectly tagged by the machine-learned model are identified. A feature vector representative of the target document is generated. Each entry of the feature vector is representative of a characteristic of the target document without including private information from the target document. The document management system queries a corpus of documents using the feature vector to identify a set of documents that correspond to the feature vector. A second training set of tagged documents is generated using the identified set of documents. The document management system retrains the machine-learned model using the second training set of tagged documents.