Fusion Deep Learning Model for Document Object Tagging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current object-recognition systems, relying on either rule-based approaches or machine-learning models, face limitations in accurately tagging document features, particularly when these features deviate from predefined rules or when insufficient training data is available, leading to inaccurate tagging and the need for costly manual updates.
Innovation Solution
A fusion deep learning model is applied to electronic documents, integrating object-recognition approaches that combine machine-learning models with user-defined rules, using feature maps and heat maps to compute tags that identify document objects, thereby leveraging the strengths of both techniques and minimizing their individual limitations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a rule-based expert system is used for object recognition, then precise tagging based on predefined document features is achieved, but the system fails to recognize document features that deviate from user-defined rules and requires costly manual updates
Solution Approach 1:
The patent combines rule-based expert systems with machine learning models into a hybrid architecture. The expert system provides precise tagging for documents matching predefined rules, while the machine learning component handles diverse and novel document features, resolving the contradiction between precision and adaptability
Solution Approach 2:
The patent introduces an intermediary layer that bridges the rule-based system and machine learning model. This intermediary processes document features and routes them to appropriate processing paths, enabling the system to maintain precision for rule-matching cases while adapting to novel features through machine learning
2Adaptability or versatility
If a machine-learning model is used for object recognition, then a wide variety of object types can be identified, but insufficient training data causes failure to produce precise tags for unseen document features
Solution Approach 1:
The patent applies preliminary rule-based processing to document features before they reach the machine learning model. This preliminary action filters and pre-processes data, allowing the machine learning model to focus on novel patterns while relying on established rules for common cases, improving precision without sacrificing versatility
Solution Approach 2:
The patent replaces pure machine learning with a hybrid system that incorporates rule-based expert systems. This substitution provides a deterministic foundation for precise tagging while retaining machine learning's ability to handle diverse document types through the combination of both approaches
3Adaptability or versatility
If manual updates are performed to maintain expert systems, then the system can adapt to changing document uses, but the process becomes costly and inefficient
Solution Approach 1:
The patent enables the system to self-update by incorporating machine learning components that automatically learn from new document data. This self-service capability reduces the need for costly manual updates while maintaining adaptability to changing document uses, significantly improving maintenance efficiency
Data Source
AI summary
Certain embodiments involve transforming an electronic document into a tagged electronic document. For instance, an electronic document processing application generates a tagged electronic document from an input electronic document. The electronic document processing application accesses one or more feature maps that identify, via a set of object-recognition rules, identified objects in the electronic document. The electronic document processing application also obtains a heat map of the electronic document that represents attributes in a pixel-wise manner. The electronic document processing application computes a tag by applying a fusion deep learning model to the one or more feature maps and the heat map. The electronic document processing application generates the tagged electronic document by applying the tag to the electronic document.