Patent Objective Identification Through Keyword Correlation Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models for identifying objectives in patent documents require large amounts of labeled data, significant computational resources, and result in high computation costs and latency, making them impractical for processing large volumes of documents.
Innovation Solution
A method and system that involves determining correlations and statistical significance of keywords to filter and generate features for training a machine learning model to identify objectives, reducing the need for extensive labeled data and computational resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If typical text classification models are used to identify problem descriptions, then identification accuracy is improved, but computation cost and latency increase significantly
Solution Approach 1:
The patent segments the document processing task into multiple stages: extracting problem descriptions using regex patterns and NLP techniques, separating them from technical details, and then processing them independently. This segmentation allows the system to achieve accurate identification without requiring computationally intensive full-document classification models.
Solution Approach 2:
The patent extracts problem descriptions from specific sections of patent documents (background, summary, detailed description) using targeted extraction methods rather than analyzing the entire document. This extraction approach reduces computation time while maintaining identification accuracy by focusing only on relevant portions.
2Reliability
If supervised machine learning models are trained to identify key descriptions, then identification capability is improved, but large amounts of labeled data and computational resources are required
Solution Approach 1:
The patent employs unsupervised learning techniques and automated rule-based systems that do not require manually labeled training data. The system automatically identifies problem descriptions using pattern matching, statistical analysis, and NLP methods, eliminating the need for expensive labeled datasets and complex supervised training pipelines.
Solution Approach 2:
The patent replaces complex supervised machine learning models with simpler rule-based systems, statistical methods, and pattern recognition algorithms. This substitution maintains identification capability while significantly reducing computational resource requirements and system complexity.
Data Source
AI summary
A method and a system for identifying an objective from documents is disclosed. In some embodiment, the method includes determining a correlation of each of a plurality of keywords extracted from a set of documents with respect to each class within a set of predefined classes. The method further includes determining a first set of keywords from the plurality of keywords. The method further includes identifying a set of data samples comprising a first plurality of sentences and a second plurality of sentences. The method further includes computing a statistical significance value of each keyword in the first set of keywords with respect to the first plurality of sentences. The method further includes generating a first set of features by discarding at least one keyword from the first set of keywords. The method further includes training a machine learning model to identify an objective of a document.