Natural Language Parsing via Ontological Segmentation and Iterative Refinement

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for transforming natural language expressions into formal language representations in computer-based information processing are limited by restrictive domains and input languages, failing to achieve the same level of accuracy as human parsing.

Innovation Solution

A method that partially parses natural language expressions by associating words with ontological entities, generating formal expressions with placeholder variables, filtering for consistency, and iteratively augmenting expressions to produce a unified formal representation suitable for downstream processing.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If severe restrictions are applied to domain or input language to achieve accurate parsing, then parsing accuracy is improved, but adaptability and versatility deteriorate

Engineering Contradiction:
Improveparsing accuracyVSAvoiddomain and input language flexibility
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The parsing process is segmented into multiple stages: initial parsing, filtering, augmentation, and iterative refinement. Each stage handles specific aspects of the parsing task, allowing the system to manage complexity systematically while maintaining high accuracy across diverse domains and input languages without requiring severe restrictions

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system implements feedback mechanisms where parsed expressions are filtered for consistency and correctness, and incorrect or incomplete parses trigger iterative refinement. This feedback loop allows the system to self-correct and improve parsing accuracy while adapting to various domains and input languages without pre-defined restrictions

Inventive Principle:
Principle #23Feedback

2Measurement precision

If multiple ontological entities are associated with each word or phrase, then parsing accuracy is improved, but device complexity increases

Engineering Contradiction:
Improveparsing accuracyVSAvoidontology management complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

Ontological entities are pre-defined and organized in a hierarchical structure before parsing occurs. This preliminary organization allows the system to efficiently associate multiple entities with words or phrases during parsing without managing complexity in real-time, as the ontology framework is already in place to guide the association process

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

Different levels of ontological detail are applied locally based on the specific word or phrase being parsed. The system associates multiple ontological entities with words or phrases only when necessary and appropriate for the given context, rather than uniformly applying complex ontology management across all parsing operations, thus reducing overall system complexity while maintaining accuracy where needed

Inventive Principle:
Principle #3Local quality

Data Source

PatentUS8024177B2Method of transforming natural language expression into formal language representation
Publication Date: 2011.09.20 MEREDITH FAMILY REVOCABLE TRUST
  • US8024177B2 patent drawing
  • US8024177B2 patent drawing
  • US8024177B2 patent drawing

AI summary

This invention comprises a series of steps which transforms one or more natural language expressions into a single, well-formed formal language representation. Each natural language expression is partially parsed into simple fragments, each of which is then associated with one or more short formal expressions. Each formal expression is constructed in such a way as to contain one or more placeholder variables, each of which is associated with one or more attributes to constrain the types of entities that each variable can potentially represent. The resulting plurality of formal expressions is then filtered for relevance within a given context, and the surviving expressions manipulated based upon a plurality of rules, which are cognizant of the attributes associated with each variable contained therein. A user is then presented with the resulting plurality of formal expressions, whereupon the user optionally selects, rejects, adds to, logically connects and otherwise manipulates each member of said plurality. When the user is satisfied that the plurality represents an intended meaning, the formal expressions are combined into a single, formal representation.