Neural Network Text Categorization Using Neighbor-Imbuing Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Computing devices face difficulties in automatically identifying the textual content that is the focus of human users within structured textual content documents, such as webpages, and existing categorization mechanisms fail to effectively exchange information across different parts of a document, leading to limited semantic understanding and filtering capabilities.
Innovation Solution
The solution involves associating individual nodes of textual content with multidimensional vectors based on their content, visual features, and positional information, which are then processed by a neighbor-imbuing neural network to enhance information exchange before being categorized by a neural network, allowing for unified and informed categorization of structured textual content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If existing categorization mechanisms are used to categorize different types of textual content independently, then each categorization task can be performed separately, but information from one part of the document cannot be utilized to improve the categorization of other parts
Solution Approach 1:
The patent merges multiple independent categorization mechanisms into a unified neural network model that processes the entire document simultaneously. This allows the model to capture cross-content information exchange and contextual relationships between different parts of the document, resolving the information loss issue while maintaining manageable complexity through integrated architecture.
Solution Approach 2:
The neural network model serves multiple functions simultaneously: it categorizes different types of textual content (headings, body text, footers, advertisements) while also capturing contextual relationships and semantic meaning across the entire document. This multi-functional approach enables a single model to handle various categorization tasks without requiring separate specialized mechanisms.
2Reliability
If traditional categorization methods are used, then the system structure remains simple, but semantic understanding and filtering capabilities are limited
Solution Approach 1:
The patent transforms the categorization approach by changing from simple rule-based classification to neural networks that process high-dimensional vector representations of text. This parameter transformation enables the system to capture subtle semantic relationships and contextual nuances, significantly improving semantic understanding accuracy while the complexity is managed through automated training processes.
3Adaptability or versatility
If large if/else branches of computer executable instructions are used for categorization, then the categorization logic can be explicitly defined, but the representational power for semantically understanding textual content is limited
Solution Approach 1:
The patent replaces the mechanical if/else branching logic with a neural network-based semantic representation system. This substitution enables the system to handle complex semantic relationships and contextual nuances that are difficult to encode in explicit rules, while the neural network's training process automatically learns optimal representations without requiring complex hand-crafted instruction structures.
Data Source
AI summary
To provide automated categorization of structured textual content individual nodes of textual content, from a document object model encapsulation of the structured textual content, have a multidimensional vector associated with them, where the values of the various dimensions of the multidimensional vector are based on the textual content in the corresponding node, the visual features applied or associated with the textual content of the corresponding node, and positional information of the textual content of the corresponding node. The multidimensional vectors are input to a neighbor-imbuing neural network. The enhanced multidimensional vectors output by the neighbor-imbuing neural network are then be provided to a categorization neural network. The resulting output can be in the form of multidimensional vectors whose dimensionality is proportional to categories into which the structured textual content is to be categorized. A weighted merge takes into account multiple nodes that are grouped together.