Text Generation Apparatus Length-Constrained Importance Estimation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional sentence summarization technologies based on neural networks fail to consider the output length when evaluating the importance of sentences or words in the input text.

Innovation Solution

A sentence generation device that includes an estimation unit to evaluate the importance of each word in the input sentence using a pre-trained model, and a generation unit to generate a summary sentence based on these importance estimates, while considering a designated output length.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional summarization technology evaluates importance of words without considering output length, then word-level importance can be estimated, but the output sentence length cannot be controlled

Engineering Contradiction:

Improveimportance evaluation precisionVSAvoidoutput sentence length

Core Design Contradiction:

Measurement precisionVSLength of moving object

Solution Approach 1:

The patent segments the summarization process into two distinct units: an estimation unit that evaluates word importance independently of length, and a generation unit that uses both the importance estimates and length information to control output. This segmentation allows each unit to specialize - the estimation unit focuses on accuracy while the generation unit handles length control.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces an intermediary mechanism where the estimation unit's importance scores serve as a bridge between the input text and the length-controlled generation process. The generation unit uses these intermediate importance estimates alongside length constraints to selectively include words in the final summary.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Ease of manufacture

If output length is not considered during importance evaluation, then importance estimation can be performed, but summaries cannot adhere to specified length constraints

Engineering Contradiction:

Improveimportance estimation simplicityVSAvoidoutput length control precision

Core Design Contradiction:

Ease of manufactureVSManufacturing precision

Solution Approach 1:

The patent performs preliminary importance estimation before length control is applied. The estimation unit first evaluates all words' importance scores without length constraints, then the generation unit uses these pre-computed scores alongside length information to select and rank words for the final summary.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent creates a dynamic two-stage process where the first stage (estimation) operates independently of length constraints, and the second stage (generation) dynamically combines the estimation results with length information to adaptively control output length while maintaining importance-based word selection.

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS12292904B2Text generation apparatus, text generation method, text generation learning apparatus, text generation learning method and program

Publication Date: 2025.05.06 NIPPON TELEGRAPH & TELEPHONE CORP

AI summary

A sentence generation device has: an estimation unit for receiving input of a first sentence and an output length, and estimating importance of each word constituting the first sentence using a pre-trained model; and a generation unit for generating a second sentence based on the importance, and thus makes it possible to evaluate importance of a constituent element of an input sentence, in correspondence with a designated output length.