Semantic Similarity Evaluation via Vector Compression Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine translation systems fail to accurately evaluate semantic similarity between sentences with adjusted word orders, as they rely on word-based evaluations that do not consider order, leading to inaccuracies.
Innovation Solution
A method and system that perform word vectorization and compression coding on sentences to generate statement vectors, calculating their distance to evaluate semantic similarity, allowing for accurate evaluation of sentences with adjusted word orders by representing them as a single vector.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If word-based semantic evaluation is used by querying external resources, then the evaluation process is simple, but the evaluation accuracy deteriorates because word order is not considered
Solution Approach 1:
The patent merges multiple word vectors into a single sentence vector through compression coding, combining both semantic meaning and word order information into one integrated representation. This resolves the contradiction by maintaining evaluation simplicity while improving accuracy through the merged vector that captures both aspects.
Solution Approach 2:
The patent changes the parameter representation from individual word vectors to a compressed sentence vector that encodes both semantic and structural information. By transforming the representation parameters, the system achieves higher evaluation accuracy without significantly increasing process complexity.
2Device complexity
If separate word comparison is performed using dictionaries, then the computational complexity is low, but the semantic similarity evaluation becomes inaccurate for sentences with adjusted word orders
Solution Approach 1:
The patent combines multiple word vectors into a single sentence vector through compression coding, integrating both semantic meaning and word order information. This approach maintains relatively low computational complexity while significantly improving evaluation accuracy for sentences with adjusted word orders.
Solution Approach 2:
The patent replaces the mechanical dictionary-based word-by-word comparison system with a neural network-based vector compression system. This substitution enables the system to capture semantic relationships and word order patterns more effectively, improving accuracy while maintaining computational efficiency.
3Measurement precision
If word vectors are compressed into statement vectors using neural networks, then the semantic similarity evaluation accuracy improves, but the device complexity increases
Solution Approach 1:
The patent merges multiple word vectors into a single sentence vector through compression coding, integrating both semantic meaning and word order information. This merging approach achieves high evaluation accuracy while keeping the system architecture relatively simple and unified.
Solution Approach 2:
The patent creates a universal sentence vector representation that serves multiple functions: capturing semantic meaning, encoding word order, and enabling similarity evaluation. This multi-functional representation improves accuracy without requiring separate systems for different evaluation aspects.
Data Source
AI summary
A semantic similarity evaluation method includes performing word vectorization processing separately on words in a first sentence and a word in a second sentence to obtain a first word vector and a second word vector; performing, in a preset word vector compression order, compression coding processing on the first word vector according to a first compression coding parameter to obtain a first statement vector; performing, in the preset word vector compression order, compression coding processing on the second word vector according to a second compression coding parameter to obtain a second statement vector; and determining a vector distance between the first statement vector and the second statement vector, and evaluating a semantic similarity between the first sentence and the second sentence according to the vector distance. The method is used to evaluate a semantic similarity.