AI Chart De-Rendering with Separate Meta and Data Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional chart de-rendering methods, both rule-based and generative, face challenges in scalability and efficiency due to complex data representation and high token usage, particularly with irregular tokenization of numbers and lack of distinction between meta and data information, leading to format errors and reduced inference speed.
Innovation Solution
A system utilizing an AI model with an image encoder and decoders to separate meta and data information, employing singularized number embeddings and multi-layer perceptrons to tokenize data efficiently, reducing token usage and minimizing format errors by representing data independently for each entity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a conventional generative model includes meta information and data information in a single data format, then scalability is improved, but data representation complexity increases and format errors occur
Solution Approach 1:
The patent segments the unified data format into separate meta information and data information formats. Meta information (chart title, axis names, legend names) is extracted and stored in a structured format with field names and types, while data information is stored separately with entity names and numerical values. This segmentation reduces representation complexity and eliminates format errors while preserving scalability through the modular structure.
2Adaptability or versatility
If numbers are treated as text and divided into irregular units, then language model compatibility is improved, but token usage increases and inference speed decreases
Solution Approach 1:
The patent changes the parameter representation of numerical data by introducing a dedicated numerical value field that stores numbers in their native format rather than as tokenized text. This allows the system to maintain compatibility with language models for processing while using efficient numerical storage and processing methods, thereby improving inference speed without sacrificing compatibility.
3Measurement precision
If a rule-based model is used for chart de-rendering, then extraction accuracy is improved, but scalability deteriorates due to requirement of separate models for each chart type
Solution Approach 1:
The patent segments the chart de-rendering process into distinct functional components: meta information extraction (handling chart title, axis names, legend names), data information extraction (handling numerical values), and structured output generation. This segmentation allows a single unified model to handle multiple chart types effectively, improving scalability while maintaining the accuracy benefits of specialized processing for each information type.
Data Source
AI summary
Provided is a system for implementing an artificial intelligence (AI) model for extracting meta information and data information included in a chart. The system includes at least one processor; and at least one memory storing instructions for the processor. The processor is configured to input the chart into an image encoder to convert the chart into a first embedding processable by the AI model, input the first embedding to the AI model to output a second embedding including the meta information from the first embedding, and to output a fourth embedding including the data information from a third embedding including information about an entity included in the second embedding, and output each of a first data format in which the meta information included in the second embedding is recorded, and a second data format in which the data information included in the fourth embedding is recorded.