Part Sorting System Using NLP Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for sorting and categorizing parts in commercial airplanes are tedious, time-consuming, and subjective, relying on human operators to review computer-aided designs, which leads to inefficiencies in managing inventories and procurement due to the large number of parts and raw text complexity in part lists.
Innovation Solution
A computer system uses natural language processing to generate sets of unigrams from part lists, creates a document term matrix, removes unigrams with occurrences above a common design threshold, and clusters parts into groups based on the processed matrix, thereby reducing noise and processing resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If human operators review computer-aided designs to sort and categorize parts, then accuracy in part grouping can be maintained, but the process becomes tedious and time-consuming due to the large number of parts
Solution Approach 1:
The patent replaces manual human review of computer-aided designs with an automated computer system that uses natural language processing to extract components from part lists and cluster parts into groups. This substitution eliminates the tedious manual sorting process while maintaining systematic accuracy through algorithmic analysis of part compositions.
Solution Approach 2:
The system enables self-service part analysis by automatically processing part lists, extracting component information, and performing clustering without requiring human operators. The computer system independently completes the entire part sorting and categorization process, freeing human resources for higher-value tasks.
2Loss of information
If raw text from part lists is used for part analysis, then complete part information is available, but the text complexity makes analysis difficult and time-consuming
Solution Approach 1:
The patent extracts specific component information from the raw text of part lists using natural language processing. By identifying and isolating individual components mentioned in part lists, the system transforms unstructured text into structured component data, reducing analysis time while preserving complete part information.
Solution Approach 2:
The system segments the raw text analysis into discrete components by extracting each component mentioned in part lists and representing it as a separate unigram. This segmentation allows the system to process complex text efficiently by breaking it down into manageable units that can be systematically analyzed and clustered.
3Measurement precision
If all components from part lists are used in analysis, then comprehensive part characterization is achieved, but processing resources and noise increase due to common design elements
Solution Approach 1:
The patent extracts and identifies common design elements (unigrams occurring above a threshold) and separates them from unique identifying features. By removing these common elements through the thresholding process, the system reduces processing complexity and noise while preserving the accuracy needed for part differentiation through the remaining unique features.
Data Source
AI summary
A method, apparatus, and computer system for grouping parts. Sets of unigrams are generated from text in part lists for parts in a family of parts using natural language processing, wherein a set of unigrams in the sets of unigrams represent components for a part in the parts. A document term matrix is created using the sets of unigrams. The document term matrix describes a presence of components in the parts. A number of unigrams from the document term matrix that has occurrences in the document term matrix that are greater than a common design threshold is removed. Removing the number of unigrams from the document term matrix forms a processed document term matrix. The common design threshold identifies a level of occurrence not useful in differentiating the parts from each other. The parts are clustered into groups using the processed document term matrix.