Document Distinguishing via LSTM Page Sequence Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies are inaccurate in distinguishing which pages belong to which documents, leading to user experience and model building inefficiencies, as they fail to consider the sequential context of pages and require tedious manual input for feature engineering and data labeling.
Innovation Solution
A computer-implemented method using a Long Short Term Memory (LSTM) model to derive relationship information between pages, automatically determining whether a page belongs to the same document or a different document by sequentially encoding page characteristics, thereby improving accuracy and eliminating the need for manual feature engineering and data labeling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing machine learning models are used to determine page-document relationships, then the system can process documents automatically, but the accuracy of document distinction is poor
Solution Approach 1:
The patent segments the document processing task into multiple independent feature extraction components (text features, layout features, image features) that are processed separately and then combined. This segmentation allows each component to be optimized independently, improving overall accuracy in determining page-document relationships while maintaining reliable automated processing.
2Measurement precision
If manual feature engineering and data labeling are performed, then the model can achieve better accuracy, but the process becomes tedious and time-consuming
Solution Approach 1:
The patent implements self-service by enabling the system to automatically extract features from documents and perform self-labeling through the trained machine learning model. The model learns to identify document boundaries and page relationships autonomously without requiring manual feature engineering or tedious data labeling, thereby achieving both high accuracy and time efficiency.
3Productivity
If sequential context of pages is not considered, then the processing is simpler and faster, but the user experience and model building experience are negatively affected
Solution Approach 1:
The patent applies preliminary action by extracting and encoding sequential context features from pages before the main classification process. The machine learning model is pre-trained to recognize sequential patterns and relationships between consecutive pages, allowing the system to maintain both high processing speed and excellent user experience by having the sequential analysis prepared in advance.
Data Source
AI summary
The accuracy of existing machine learning models, software technologies, and computers are improved by estimating whether a particular page belongs to a same document as another page or whether the page belongs to a different document. Such document distinguishing can be based on deriving relationship information between a first feature vector representing the page and a second feature vector representing the other page. This also improves the user experience and model building experience, among other things.