Word-Level Sentence Comparison for Meaning-Aware Edit Distance

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing sentence comparison tools, such as 'COPIPERIN', are inadequate for grasping the meaning of entire sentences due to unsuitable comparison algorithms and display manners, making it difficult to determine similarities and differences accurately.

Innovation Solution

A sentence management system that calculates sentence distance by counting edits on a per-word basis and displays the results in a manner that highlights edits and similarities, using edit weights to enhance meaning grasping, and employs triangle inequality to optimize calculations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a sentence comparison tool uses a comparison algorithm designed for determining copied and pasted text, then it can identify text duplication, but it cannot accurately grasp the meaning of entire sentences

Engineering Contradiction:
Improveaccuracy in determining copied and pasted textVSAvoidloss of sentence meaning
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

The patent segments sentences into individual words or phrases and compares them separately. The comparison unit is divided into a query sentence and multiple candidate sentences, with each candidate's similarity calculated based on word-level matching. This segmentation allows the system to preserve sentence meaning while identifying duplication, as it can distinguish between exact copies and sentences with different meanings but similar wording.

Inventive Principle:
Principle #1Segmentation

2Productivity

If a sentence comparison tool displays only similarity percentages, then it provides quick comparison results, but it does not help users understand the differences between sentences

Engineering Contradiction:
Improvespeed of comparisonVSAvoidloss of difference information
Core Design Contradiction:
ProductivityVSLoss of information

Solution Approach 1:

The patent applies local quality by highlighting specific portions of sentences rather than treating them uniformly. Different display styles are applied to matched portions versus unmatched portions, allowing users to quickly identify where sentences differ. This localized differentiation preserves both the speed of comparison and the information about differences, as users can visually locate dissimilarities without reading entire sentences.

Inventive Principle:
Principle #3Local quality

3Reliability

If a sentence comparison tool compares every candidate sentence with the query sentence, then it ensures comprehensive coverage, but it increases computational load

Engineering Contradiction:
Improvecompleteness of comparisonVSAvoidcomputational energy
Core Design Contradiction:
ReliabilityVSLoss of energy

Solution Approach 1:

The patent implements partial action by introducing a threshold mechanism that stops comparison early when sufficient results are found. If a candidate sentence's similarity exceeds a predetermined threshold, the system can terminate further comparisons or reduce the search scope. This approach maintains reliability by ensuring that sentences above the threshold are identified, while reducing computational energy by avoiding unnecessary comparisons with clearly dissimilar sentences.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentUS12561516B2Sentence management system and storage medium storing sentence management program
Publication Date: 2026.02.24 SEIKO EPSON CORP
  • US12561516B2 patent drawing
  • US12561516B2 patent drawing
  • US12561516B2 patent drawing

AI summary

A sentence management system includes an obtainer configured to obtain a sentence of interest and a plurality of comparison sentences, a distance calculator configured to calculate a sentence distance between the sentence of interest and each of the comparison sentences, and a display processor configured to cause the sentence of interest, the comparison sentences, and a list of comparison sentence names arranged in order of the sentence distance to be displayed simultaneously. The distance calculator is configured to calculate the sentence distance by calculating a number of edits on a per-word basis between the sentence of interest and each of the comparison sentences.