Iterative Dictionary Expansion for NLP Accuracy

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing natural language processing systems face challenges in efficiently identifying and understanding dynamic structures and changes in semantic relationships, leading to inaccurate outcomes when processing natural language and extracting information from unstructured text.

Innovation Solution

An AI platform with a dictionary manager, director, and text manager is used to iteratively expand a domain-specific dictionary by adding semantically similar and contextually related instances, allowing for improved identification of matching corpus data through the application of expanded dictionaries to unexplored corpora.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a static domain-specific dictionary is used for information extraction, then the system structure is simple and easy to maintain, but the accuracy and relevance of information extraction deteriorates due to inability to capture dynamic semantic relationships

Engineering Contradiction:

Improveaccuracy of information extractionVSAvoiddictionary expansion system complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent implements an iterative dictionary expansion system that dynamically updates the domain-specific dictionary by identifying and adding semantically similar instances and extension instances. The director component continuously expands the dictionary based on unexplored corpora, transforming the static dictionary into a dynamic structure that adapts to new semantic relationships and contextual variations, thereby improving information extraction accuracy without requiring complete system redesign

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The dictionary expansion system operates autonomously through the director component, which automatically identifies semantically similar instances using neural language models and adds relevant terms to the dictionary without requiring manual intervention. The system self-updates by processing unexplored corpora and expanding the dictionary instances based on identified semantic relationships, reducing the need for external maintenance while improving extraction accuracy

Inventive Principle:

Principle #25Self-service

2Adaptability or versatility

If the dictionary is expanded manually to include more terms, then the coverage of semantic relationships improves, but the time and resources required for maintenance increases

Engineering Contradiction:

Improvesemantic relationship coverageVSAvoiddictionary maintenance time

Core Design Contradiction:

Adaptability or versatilityVSLoss of time

Solution Approach 1:

The system implements a feedback mechanism where the director continuously evaluates unexplored corpora against the current dictionary, identifies gaps in semantic coverage, and automatically expands the dictionary to fill these gaps. The iterative process uses feedback from neural language model analyses to guide dictionary expansion, ensuring that semantic relationship coverage improves systematically while maintaining automated efficiency that reduces manual maintenance time

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system performs preliminary expansion actions by pre-identifying semantically similar instances and extension instances from unexplored corpora before they are needed for information extraction. The director proactively expands the dictionary in advance based on anticipated semantic needs, reducing the time required for manual updates by automating the preliminary work of identifying and adding relevant terms before they are actually required for extraction tasks

Inventive Principle:

Principle #10Preliminary action

3Measurement precision

If neural language models are used to identify semantically similar instances, then the precision of semantic matching improves, but the computational resources and processing time increase

Engineering Contradiction:

Improvesemantic similarity identification accuracyVSAvoidcomputational resource consumption

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The system applies partial action by using neural language models selectively only for identifying semantically similar instances and extension instances during dictionary expansion, rather than applying them to every information extraction task. The director uses neural models to analyze unexplored corpora and expand the dictionary, but once expanded, the system uses the enriched dictionary for efficient matching without requiring continuous neural model invocation, thus achieving high semantic precision while reducing overall computational resource consumption

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS11030402B2Dictionary expansion using neural language models

Publication Date: 2021.06.08 INTERNATIONAL BUSINESS MACHINE CORPORATION

AI summary

Embodiments relate to a system, program product, and method for iterative expansion and application of a domain-specific dictionary. One or more dictionary instances are applied against a text corpus. The dictionary is iteratively expanded and selectively populated with one or more additional dictionary instances, including semantically similar instances to the applied dictionary instances and extension instances contextually related to the applied dictionary instances. The iteratively expanded dictionary is applied to an unexplored corpus to identify matching corpus data to populated instances of the dictionary.