Morpheme-Based Polysemous Word Vector Assignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional techniques struggle to assign appropriate vectors to polysemous words, leading to reduced precision in sentence analysis and translation, especially when dealing with words having multiple meanings and idiomatic expressions.
Innovation Solution
A computer-readable recording medium that processes input text by performing morpheme analysis to identify polysemous words and their semantic divisions, using co-occurrence rate tables and category classifications to assign specific vectors to each meaning of the polysemous words, thereby enhancing the precision of sentence vectors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If the same vector is assigned to all meanings of a polysemous word, then the system is simple to operate, but the precision of sentence analysis deteriorates
Solution Approach 1:
The patent segments polysemous words into multiple semantic divisions (e.g., cool(1) for personality, cool(2) for temperature). Each semantic division is assigned a separate vector, allowing the system to distinguish between different meanings of the same word. This segmentation resolves the contradiction by maintaining simplicity in the assignment process while improving analysis precision through meaning-specific vectors.
Solution Approach 2:
The patent applies local quality by assigning different vector characteristics to different semantic divisions of polysemous words based on their specific meanings. Instead of using a uniform vector for all meanings, the system tailors the vector properties to local semantic requirements, thereby improving precision without significantly complicating the overall system operation.
2Measurement precision
If HMM parameters are set with all semantic divisions of polysemous words, then the precision of vector assignment improves, but the device complexity increases due to huge data registration
Solution Approach 1:
The patent extracts only the essential co-occurrence information between polysemous words and their neighboring words, storing this in a compact co-occurrence rate table. This extraction approach captures the necessary semantic distinctions without requiring complete HMM parameter registration for all possible semantic divisions, thereby reducing device complexity while maintaining precision.
Solution Approach 2:
The patent performs preliminary action by pre-calculating and storing co-occurrence rates between polysemous words and their context words in a dedicated table. This pre-computation allows the system to quickly identify semantic divisions during text processing without requiring complex real-time HMM parameter calculations, thus reducing overall system complexity.
3Quantity of substance
If conventional word vector techniques are used, then the data amount is small, but the precision of analyzing polysemous words deteriorates
Solution Approach 1:
The patent segments polysemous words into multiple semantic divisions and assigns specific vectors to each division. This segmentation enables the system to use minimal additional data (only co-occurrence rates between polysemous words and context words) to achieve significant improvements in analysis precision, avoiding the need for large-scale data registration while maintaining data efficiency.
Data Source
AI summary
A non-transitory computer-readable recording medium has stored therein an information processing program that causes a computer to execute a process including separating input text into a plurality of words, by performing a morpheme analysis on the input text identifying, from among the plurality of words contained in the input text, a polysemous word and a semantic division of the polysemous word, based on a co-occurrence rate table defining co-occurrence rates between polysemous words distinguished according to semantic divisions thereof and categories of words and based on the plurality of words contained in the input text, and assigning a vector corresponding to the semantic division of the polysemous word, to the polysemous word contained in the input text.