ML Synthesized Document Generation via Keyword Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current information processing systems face inefficiencies in managing and analyzing unstructured text data, requiring manual screening and rule customization, which is tedious and time-consuming, especially for large volumes.
Innovation Solution
A machine learning-based approach that generates synthesized documents by extracting keywords from multiple sections of documents using BERT and Seq2Seq models, ensuring context continuity and automating the content selection process based on search terms and similarity analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual screening and rule customization are used to process unstructured text data, then processing accuracy can be maintained, but processing time and labor requirements increase significantly
Solution Approach 1:
The patent replaces manual screening and rule customization with machine learning models (BERT for keyword extraction, Seq2Seq for content generation) that automatically process unstructured text data, eliminating the need for human operators to manually review and sample documents while maintaining processing accuracy through learned patterns and similarities
2Measurement precision
If manual customization of rules is performed to determine correspondence with predefined themes, then theme accuracy can be ensured, but operational complexity and time consumption increase
Solution Approach 1:
The system performs self-service by automatically extracting keywords from documents using BERT models and generating synthesized content using Seq2Seq models without requiring manual rule customization. The model learns theme correspondence patterns from training data and applies them autonomously, eliminating the need for operators to create and maintain complex rule sets
3Productivity
If automated keyword extraction is implemented, then processing speed improves, but the complexity of the system increases
Solution Approach 1:
The patent segments the document processing task into distinct functional components: BERT models for keyword extraction from individual documents, Seq2Seq models for generating synthesized content, and a similarity calculation module for matching keywords with search terms. This segmentation allows each component to be optimized independently while working together to achieve high processing speed
4Manufacturing precision
If multiple machine learning models are used for content selection and generation, then synthesized document quality improves, but computational resources and system complexity increase
Solution Approach 1:
The system performs preliminary action by pre-training BERT models for keyword extraction and Seq2Seq models for content generation on large corpora of documents before actual processing. This pre-training allows the models to make accurate predictions and generate high-quality synthesized documents during inference without requiring excessive computational resources, as the heavy lifting is done during the initial training phase
Data Source
AI summary
An apparatus comprises a processing device configured to receive a request to generate a synthesized document comprising one or more search terms, and to extract, utilizing a first machine learning model, keywords from a set of documents. The processing device is also configured to select first content for inclusion in a first section of the synthesized document based on a similarity of the search terms and the extracted keywords from corresponding first sections of the set of documents, and to determine, utilizing a second machine learning model that takes as input the selected first content, a set of terms for a second section of the synthesized document. The processing device is further configured to select second content for inclusion in the second section of the synthesized document based on a similarity of the determined set of terms and the extracted keywords from corresponding sections of the set of documents.