Graph Edge Vector Generation for Extensible Text Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing text recognition models have low extensibility and inefficient parameter update processes due to the reliance on initial edge vectors generated solely from initial node vectors, consuming significant computing resources and reducing model efficiency.
Innovation Solution
A method involving generating initial node vectors and edge vectors based on associated node edges, using general training parameters to adjust model parameters, thereby improving model extensibility and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the initial edge vector is generated only by using the initial node vectors, then the model structure is simple, but the extensibility of the text recognition model is relatively low
Solution Approach 1:
The edge vector generation process is segmented into multiple independent components: initial edge vector generation from node vectors, association strength calculation between entities, and final edge vector construction combining both elements. This segmentation allows each component to be optimized independently, improving extensibility without overwhelming complexity.
Solution Approach 2:
The model transitions from a static edge vector generation method to a dynamic one where association strengths are calculated based on entity relationships and used to weight different vector components. This dynamic approach enables the model to adapt to different relationship types and scenarios, significantly improving extensibility.
2Productivity
If the initial edge vector is inferred before model parameter adjustment, then the model structure is complete, but the computing resources consumed are large and model parameter update efficiency is reduced
Solution Approach 1:
The initial edge vector is pre-calculated from node vectors before the training process begins. This preliminary action allows the main training loop to focus only on updating association strengths and final edge vectors, significantly reducing the computational burden during parameter updates and improving overall efficiency.
Solution Approach 2:
Instead of recalculating all edge vectors during each parameter update, the model updates only the association strength parameters locally based on gradient descent. This localized update approach reduces computing resource consumption while maintaining model completeness and accuracy.
Data Source
AI summary
A data processing method including obtaining training sample set including sample text, tag entity words in the sample text, and tag entity word pair carrying tag relationship, inputting the training sample set into an initial text recognition model to generate initial node vectors respectively corresponding to nodes including sentence nodes and entity nodes, generating an associated node edge according to the nodes, determining, according to general training parameter of the associated node edge and initial node vectors corresponding to connected nodes connected by the associated node edge, an initial edge vector corresponding to the associated node edge, generating an initial graph structure including initial node vectors and the initial edge vector, and adjusting parameter in the initial text recognition model according to initial node vectors corresponding to every two entity nodes, the tag entity word pair, and the initial graph structure, to obtain text recognition model.