Text Similarity Calculation Using Vector Codes and Fusion Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing text similarity calculation methods fail to accurately determine similarity between texts with the same ideograms but different characters, as they do not consider semantic information effectively.
Innovation Solution
A method that converts texts into vector codes, obtains fusion features, and calculates semantic features based on character string and coding sequences to determine text similarity, incorporating neural networks and attention mechanisms for comprehensive feature representation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If text similarity is calculated based on character repetition degree, then calculation simplicity is maintained, but recognition accuracy deteriorates when texts have same ideograms but different characters
Solution Approach 1:
The patent segments text similarity calculation into multiple independent feature dimensions: character string features (based on repetition degree), coding sequence features (vector representations), and fusion features. Each dimension captures different aspects of text similarity, allowing the system to maintain calculation simplicity in the character level while achieving high accuracy through multi-dimensional feature integration
Solution Approach 2:
The patent creates a composite feature representation by combining character string features, coding sequence features, and their fusion features. This composite approach integrates both simple character-level matching and complex semantic-level representations, resolving the contradiction between calculation simplicity and recognition accuracy by leveraging the strengths of different feature types
2Use of energy by moving object
If traditional character repetition methods are used, then computational resources are saved, but semantic information is not effectively considered
Solution Approach 1:
The patent transforms text representation from traditional character-level one-dimensional analysis to multi-dimensional feature space by introducing coding sequences (vector codes) and fusion features. This dimensional expansion enables the system to capture semantic information that is invisible to traditional character repetition methods, while the hierarchical feature extraction process efficiently manages computational resources by building upon simpler character features
3Measurement precision
If multi-feature fusion is implemented, then text similarity accuracy is improved, but system complexity increases
Solution Approach 1:
The patent performs preliminary extraction of character string features and coding sequence features separately before fusing them. This preliminary action organizes complex information into structured intermediate representations, making the subsequent fusion process more manageable and systematic, thereby improving text similarity accuracy while controlling system complexity through structured feature engineering
Data Source
AI summary
A method for determining text similarity, a storage medium and an electronic device. The method includes: obtaining a first coding sequence corresponding to the first text and a second coding sequence corresponding to the second text; obtaining a first fusion feature corresponding to the first coding sequence and a second fusion feature corresponding to the second coding sequence based on the first coding sequence and the second coding sequence; converting the first fusion feature and the second fusion feature into a corresponding first semantic feature and a second semantic feature for representing semantics, and determining a text similarity between the first text and the second text based on the first semantic feature and the second semantic feature.