Word Semantic Relation Estimation Using Combined Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating semantic relations between words, particularly antonyms, are inefficient due to sparse similarity calculations and invalid feature vectors, which hinder accurate discrimination of meaning differences.
Innovation Solution
A word semantic relation estimation device that combines word vectors and character vectors to generate a valid feature amount, allowing for improved estimation of semantic relations between words, including antonyms, by using a database search unit, word vector extracting unit, character vector converting unit, and estimation unit configured with supervised learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If word vectors are used to represent word meanings, then semantic similarity can be calculated, but antonyms with different meanings appearing in similar contexts are converted into similar vectors, causing incorrect semantic relation estimation
Solution Approach 1:
The patent segments the word representation into two parts: word vector (capturing contextual meaning) and character vector (capturing lexical form). This segmentation allows the system to separately consider both contextual similarity and lexical differences, resolving the contradiction where antonyms appear similar in context but differ in meaning.
Solution Approach 2:
The patent creates a composite feature vector by combining word vectors and character vectors. This composite representation integrates both contextual information (from word vectors) and lexical information (from character vectors), enabling accurate distinction between antonyms that have similar contexts but opposite meanings.
2Adaptability or versatility
If multiple types of similarities are calculated for word pairs, then comprehensive feature vectors can be generated, but the calculation method is sparse and the feature vectors may not be valid
Solution Approach 1:
The patent changes the parameters used for feature representation from sparse similarity calculations to dense vector representations (word vectors and character vectors). This parameter change transforms the feature space from sparse and potentially invalid to dense and reliable, while maintaining comprehensiveness through the combination of multiple vector types.
Data Source
AI summary
The following units are included: a word search unit searching a database for each of words included in an input word pair; a word vector extracting unit extracting a word vector corresponding to each of the searched words from a word vector database; a character vector converting unit converting characters included in the word pair into a character vector; a feature amount acquiring unit acquiring a feature amount by combining the extracted word vector of each of the words and the character vector generated by converting the characters; and an estimation unit estimating a semantic relation between the words from the acquired feature amount.