Word Vector Conversion Using Semantic Dictionary Scaling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional word vector generation methods fail to accurately represent semantic relationships between words, leading to inappropriate distance relationships between text vectors, particularly for antonyms, co-hyponyms, and synonyms, resulting in incorrect similarity calculations between texts.
Innovation Solution
A word vector changing method that adjusts vector distances based on semantic relationships using a dictionary of semantically distant or close word pairs, where the scalar value k is set to greater than 1 for distant pairs and less than 1 for close pairs, while minimizing changes for non-listed pairs, to ensure appropriate vector-to-vector distances.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If word vectors are generated using conventional distribution hypothesis methods, then the generation process is simple and fast, but the distance relationship between semantically related words (antonyms, co-hyponyms, synonyms) becomes inaccurate
Solution Approach 1:
The patent applies preliminary action by pre-defining a dictionary of semantically related word pairs (antonyms, co-hyponyms, synonyms) before vector generation. This dictionary is used to guide the vector conversion process, ensuring that known semantic relationships are preserved in the final vector space without requiring complex real-time analysis during vector generation.
Solution Approach 2:
The patent changes parameters by introducing a conversion function that adjusts word vectors based on their semantic relationships. The function modifies vector distances by applying scaling factors to specific word pairs from the dictionary, transforming the raw distributional vectors into semantically-aware vectors that maintain appropriate distance relationships.
2Measurement precision
If the vector distance for semantically distant pairs (antonyms, co-hyponyms) is increased, then the semantic distinction becomes clearer, but the distance for other word pairs may become unreasonable
Solution Approach 1:
The patent applies local quality by treating different word pairs differently based on their semantic relationships. The conversion function selectively adjusts distances for specific categories (antonyms, co-hyponyms, synonyms) while leaving other word pairs unchanged or minimally affected. This localized adjustment ensures that semantic distinctions are enhanced where needed without distorting the overall vector space structure.
Solution Approach 2:
The patent uses feedback by iteratively adjusting word vectors based on the predefined dictionary of semantically related pairs. The conversion function continuously refines vector distances by comparing against known semantic relationships, ensuring that adjustments to enhance distinction clarity do not compromise overall distance relationship reliability.
3Measurement precision
If conventional word vector generation is used, then computational resources and time are saved, but the accuracy of text similarity calculation deteriorates due to inappropriate vector distances
Solution Approach 1:
The patent applies partial action by focusing computational efforts only on adjusting vectors for semantically related word pairs that appear in the predefined dictionary. Rather than reprocessing all word vectors or performing exhaustive semantic analysis, the system selectively converts only the necessary pairs, maintaining processing efficiency while improving text similarity accuracy.
Solution Approach 2:
The patent changes parameters by transforming the vector space through a conversion function that scales distances for specific word pairs. This parameter transformation improves measurement precision for text similarity calculations without requiring complete regeneration of all word vectors, thus preserving productivity.
Data Source
AI summary
To arrange all words so that the distance of a given word pair will be appropriate.Using as input a concept base 22 which is a set of pairs of a word and a vector representing a concept of the word, and a dictionary 24 which is a set of semantically distant or close word pairs, when a word pair C being a pair of given words A, B in the concept base 22 is present in the dictionary 24, conversion means 30 associates with the word pair C a magnitude D of a difference vector between a difference vector V′ between a converted vector of the word A and a converted vector of the word B, and a vector kV determined by multiplying a difference vector V between the vector of the word A in the concept base 22 and the vector of the word B in the concept base 22 by a scalar value k. When the word pair C is not present in the dictionary 24, the conversion means 30 associates the magnitude D of the difference vector between the difference vector V′ and the difference vector V with the word pair C. The conversion means 30 converts the vector of a given word in the concept base 22 such that a total sum of the magnitude D corresponding to every word pair C is minimized.