Natural Language Pattern Generation for Word Sense Disambiguation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computer systems fail to accurately determine the appropriate senses of words in a sentence, leading to ineffective information retrieval and search results, as seen in examples like 'the table ate the pie' and 'the dog ate the pie', where semantic meaning is misinterpreted.
Innovation Solution
A computer-implemented method and system generate 'declared patterns' from sentence components using a parts of speech tagger, extended dictionaries with emergent categories, and equivalent pattern specification sets to identify valid word senses, creating a mapping between natural language grammar and semantic items.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a simple parts of speech tagger is used to identify word categories, then the processing speed is fast, but the accuracy of determining word senses is insufficient
Solution Approach 1:
The system segments the word sense determination process into multiple stages: initial POS tagging, pattern unit generation, grammatical role identification, and sense determination. Each stage processes specific aspects separately, allowing the system to achieve high accuracy without overwhelming complexity at any single stage.
Solution Approach 2:
The patent introduces intermediate structures (pattern units, grammatical roles, and equivalence classes) that mediate between simple POS tagging and final sense determination. These intermediaries bridge the gap between basic grammatical classification and sophisticated semantic understanding.
2Reliability
If multiple dictionaries and databases are used to determine word senses, then the accuracy of information retrieval is improved, but the processing time and computational resources increase
Solution Approach 1:
The system performs preliminary organization of linguistic data by pre-establishing equivalence classes for pattern units and grammatical roles. This preliminary structuring allows faster access and processing during actual information retrieval operations, reducing processing time while maintaining accuracy.
Solution Approach 2:
The patent transforms the problem from direct sense matching to equivalence class-based pattern matching. By changing the parameter of comparison from specific word senses to broader equivalence classes, the system reduces computational complexity while preserving retrieval accuracy.
3Measurement precision
If emergent categories are introduced to determine valid word senses, then the semantic understanding is enhanced, but the complexity of the dictionary structure increases
Solution Approach 1:
The patent adds a new dimension to the dictionary structure by introducing equivalence classes that operate alongside traditional POS tags and grammatical roles. This additional dimensional organization enables sophisticated semantic understanding while maintaining the原有 simple structure through layered classification.
Data Source
AI summary
Disclosed herein is a computer implemented method and system of generating declared patterns from components of a sentence. Parts of speech in the sentence are tagged for identifying parts of speech of each word and phrase in the sentence. Sentence chunking is then performed using the identified parts of speech of each word and phrase to generate pattern units. A first dictionary and a database of equivalent pattern specification sets are then applied to identify grammatical roles and senses of the generated pattern units. A second dictionary and a conceptionary are then applied to identify an equivalent name set for each of the generated pattern units. The declared patterns are then generated for the sentence using the identified equivalent name set for each of the generated pattern units.