Automated Teacher Data Tagging via Relevance Calculation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Manual tagging of teacher data is labor-intensive and time-consuming, making it difficult to supplement appropriately imparted tags in documents for automatic test item extraction.

Innovation Solution

A teacher data generation apparatus that calculates the relevance degree between document descriptions and corresponding tags, automatically imparting tags to portions with a high relevance degree using pointwise mutual information, thereby generating teacher data for efficient learning and extraction.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If tags are manually imparted in teacher data, then tagging accuracy can be ensured, but it takes a great deal of work and time to supplement teacher data

Engineering Contradiction:

Improvetagging accuracyVSAvoidtime to supplement teacher data

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent uses pointwise mutual information to calculate the relevance degree between document descriptions and tag descriptions, automatically copying appropriate tags from the calculated relevance results. This replaces manual tagging while maintaining accuracy through quantitative measurement, resolving the contradiction between tagging accuracy and time consumption.

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The patent replaces the mechanical manual tagging process with an automated calculation system that uses pointwise mutual information to determine tag relevance. This substitution eliminates manual labor while preserving tagging quality through mathematical computation, directly addressing the time loss issue.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Reliability

If tags are manually imparted in teacher data, then appropriate tags can be supplemented, but the process becomes labor-intensive

Engineering Contradiction:

Improveappropriateness of tagsVSAvoidease of supplementing teacher data

Core Design Contradiction:

ReliabilityVSEase of manufacture

Solution Approach 1:

The system performs self-service by automatically calculating the relevance degree between document descriptions and tag descriptions using pointwise mutual information, and autonomously imparting tags without human intervention. This maintains reliability through quantitative assessment while dramatically improving ease of supplementing teacher data.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent changes the parameter of tag supplementation from manual operation to automated calculation based on relevance degree thresholds. By introducing the parameter of pointwise mutual information calculation, the system maintains tag appropriateness while transforming the process into an easily executable automated procedure.

Inventive Principle:

Principle #35Parameter changes

3Productivity

If automatic tag imparting is implemented without relevance calculation, then the process becomes fast, but tag appropriateness deteriorates

Engineering Contradiction:

Improvespeed of tag supplementationVSAvoidtag appropriateness

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent implements feedback by calculating the relevance degree between document descriptions and tag descriptions using pointwise mutual information before imparting tags. This feedback mechanism ensures tag appropriateness is maintained while the automated process preserves high productivity, resolving the contradiction between speed and quality.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS12153877B2Training data generation device, training data generation method and training data generation program

Publication Date: 2024.11.26 NIPPON TELEGRAPH & TELEPHONE CORP

AI summary

In teacher data generation processing for generating teacher data in which a tag is imparted in a document, a calculation unit (15a) calculates a relevance degree between a description in a document and a description in a document corresponding to a tag. When the calculated relevance degree is equal to or greater than a predetermined threshold, the imparting unit (15b) imparts the tag to a portion where the description is provided in the document.