Tree Structure Data Grouping Using Similarity Thresholds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for grouping similar data, such as those described in PTL 1 and NPL 1, are limited in their ability to handle general data and are inefficient due to high operation times, particularly because they require time-consuming similarity searches for each search result, resulting in processing times of O(N2).
Innovation Solution
An information processing device and method that utilizes a tree structure to search and group data based on similarity, including a search unit, grouping determination unit, subtree grouping unit, leaf node grouping unit, data merging unit, and group merging unit, which determines and merges groups based on similarity thresholds, reducing the need for repeated searches and grouping operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If similarity search is executed for each search result to group data, then grouping accuracy is improved, but processing time increases to O(N2)
Solution Approach 1:
The patent applies preliminary action by pre-constructing a tree structure that organizes data based on similarity relationships before the actual grouping operation. This tree structure pre-computes and stores similarity information, allowing the grouping process to simply traverse the pre-built structure rather than performing repeated similarity searches. The tree structure is built once beforehand, and subsequent grouping operations benefit from this pre-prepared organization, reducing time complexity from O(N2) to O(N).
Solution Approach 2:
The patent segments the data into a hierarchical tree structure where data is organized in nodes and edges represent similarity relationships. By dividing the data into manageable segments (nodes) with defined relationships (edges), the system avoids comparing all data points against each other. Instead, grouping operations traverse the segmented tree structure, significantly reducing the number of comparisons needed while maintaining grouping accuracy.
2Reliability
If repeated similarity searches and grouping operations are performed, then comprehensive grouping is achieved, but operation time increases
Solution Approach 1:
The patent performs preliminary action by constructing the tree structure in advance, which encodes all necessary similarity relationships. This pre-computation ensures that comprehensive grouping information is available without needing repeated searches. The tree structure captures the complete similarity landscape once, and subsequent grouping operations simply traverse this pre-established structure to identify groups, ensuring both completeness and efficiency.
Solution Approach 2:
The patent creates a copied representation of the data relationships in the form of a tree structure. Instead of repeatedly accessing and analyzing the original data for similarity, the system uses the tree structure as a copied model that contains all similarity information. This copying allows efficient traversal and grouping operations without repeatedly processing the original data, maintaining grouping completeness while improving operation efficiency.
3Speed
If tree structure is used to speed up similarity search, then search efficiency is improved, but grouping method is not provided
Solution Approach 1:
The patent makes the tree structure multi-functional by using it for both similarity search and grouping operations. The same tree structure that accelerates similarity search is also directly utilized to perform grouping by traversing nodes and edges. This universal application of the tree structure eliminates the need for separate grouping algorithms, providing both search efficiency and comprehensive grouping capability through a single data structure.
Solution Approach 2:
The patent merges the similarity search function and grouping function into a unified process using the tree structure. Instead of performing similarity search and then separately executing grouping operations, the system combines both functions by traversing the tree structure once to identify groups. This merging of functions maintains the search speed benefits of the tree structure while adding complete grouping capability, avoiding the need for separate processing steps.
Data Source
AI summary
An information processing device according to the present invention includes: a search unit that searches data in a tree structure having a node; a grouping determination unit that determines whether to group using data and a lower node based on a similarity associated with an edge between data and the lower node of the data and a predetermined threshold; a subtree grouping unit that generates a group by grouping the data and the lower node; a leaf node grouping unit that generates one or a plurality of groups by grouping the leaf node when the node is a leaf node; a data merging unit that merges the data into any one of groups of lower nodes of the data when data returned in back track to an upper node are not belonged a group; and a group merging unit that merges some of the groups.