Hybrid NLG System Template Selection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing natural language generation (NLG) systems require significant human investment and are not adaptable to new domains, with rule-based systems relying heavily on human expertise and statistical systems being computationally expensive and less natural in output.

Innovation Solution

A hybrid NLG system that combines statistical and template-based approaches, using a ranking support vector machine to determine optimal templates from a training corpus, reducing human-intensive rule generation and domain adaptability while eliminating the need for extensive grammatical correction.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If rule-based NLG systems are used, then high quality objective text generation is achieved, but intensive human investment and domain expertise are required

Engineering Contradiction:

Improvetext generation qualityVSAvoidhuman investment requirement

Core Design Contradiction:

Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent uses template-based generation where pre-defined templates are copied and filled with domain-specific parameters. This eliminates the need for extensive rule creation while maintaining structured output quality, as templates serve as reusable patterns that can be adapted to different domains without manual rule engineering

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The system transforms rigid rule-based generation into parameter-driven template instantiation. By changing from fixed rules to parameterized templates with statistical selection, the system maintains text quality while reducing human investment in rule creation and enabling domain adaptability through parameter adjustment rather than rule rewriting

Inventive Principle:

Principle #35Parameter changes

2Adaptability or versatility

If statistical NLG systems are used, then domain adaptability and reduced human investment are achieved, but computational expense increases and output naturalness decreases

Engineering Contradiction:

Improvedomain adaptabilityVSAvoidcomputational expense

Core Design Contradiction:

Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

The patent segments the generation process into distinct stages: template selection, parameter filling, and post-processing. This segmentation allows statistical methods to be applied only where needed (template selection) while using more efficient rule-based approaches for structured operations, reducing overall computational expense while maintaining domain adaptability

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system applies statistical methods partially - only for template selection and ranking - rather than throughout the entire generation process. This partial application of statistical approaches provides domain adaptability where most needed while avoiding excessive computational expense in stages where rule-based methods suffice

Inventive Principle:

Principle #16Partial or excessive action

3Extent of automation

If statistical NLG systems are used, then human investment is reduced, but grammatical correctness cannot be guaranteed

Engineering Contradiction:

Improvehuman investment reductionVSAvoidgrammatical correctness

Core Design Contradiction:

Extent of automationVSReliability

Solution Approach 1:

The patent merges statistical template selection with rule-based grammatical constraints. Templates provide statistically-informed structure while embedded grammatical rules ensure correctness, combining the advantages of both approaches to reduce human investment while maintaining reliability of output

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

Templates serve as intermediaries between statistical corpus data and grammatical output. The templates encode grammatical structures learned from corpora, acting as a mediator that translates statistical patterns into grammatically correct sentences without requiring direct statistical generation of each word

Inventive Principle:

Principle #24Intermediary (Mediator)

4Manufacturing precision

If template-based NLG systems are used, then grammatical correctness is improved, but text naturalness and variability decrease

Engineering Contradiction:

Improvegrammatical correctnessVSAvoidtext naturalness

Core Design Contradiction:

Manufacturing precisionVSAdaptability or versatility

Solution Approach 1:

The patent makes template selection dynamic through statistical ranking based on corpus data. Instead of using fixed templates, the system dynamically selects and ranks templates based on their suitability for the current context, enabling text naturalness and variability while maintaining grammatical correctness through the template structure

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS9953031B2Systems and methods for natural language generation

Publication Date: 2018.04.24 THOMSON REUTERS ENTERPRISE CENTRE GMBH

AI summary

A method includes receiving a corpus comprising a set of pre-segmented texts. The method further includes creating a plurality of modified pre-segmented texts for the set of pre-segmented texts by extracting a set of semantic terms for each pre-segmented text within the set of pre-segmented texts and applying at least one domain tag for each pre-segmented text within the set of pre-segmented texts. The method further includes clustering the plurality of modified pre-segmented texts into one or more conceptual units, wherein each of the one or more conceptual units is associated with one or more templates, wherein each of the one or more templates corresponds to one of the plurality of modified pre-segmented texts.