Word Mover's Similarity for Efficient Text Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing natural language processing frameworks face inefficiencies and high computational costs when performing fast listwise text comparison tasks, such as comparing a set of target text data objects to a reference text data object, due to the need for numerous document comparison actions.
Innovation Solution
The use of Word Mover's Similarity, Relaxed Word Mover's Similarity, and Related Relaxed Word Mover's Similarity measures to determine text similarity by generating a maximal word similarity score based on word-wise flow and similarity values, framing the problem as a maximization problem to reduce computational complexity and increase efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional document comparison methods are used to perform listwise text comparison, then measurement precision is maintained, but computational complexity increases and productivity decreases
Solution Approach 1:
The patent segments the document comparison task into word-level operations. Instead of comparing entire documents directly, it breaks down the comparison into individual word embeddings and their transformations, allowing parallel processing and reducing overall computational complexity while maintaining accuracy
Solution Approach 2:
The patent introduces Word Mover's Distance (WMD) as an intermediary measure between traditional document comparison and simple word matching. WMD serves as a computational bridge that provides meaningful semantic distance metrics without requiring exhaustive document-level comparisons, thus improving efficiency while preserving measurement precision
2Loss of time
If traditional document comparison methods are used, then measurement precision is maintained, but loss of time increases
Solution Approach 1:
The patent performs preliminary actions by pre-computing word embeddings and storing them in a vocabulary database before actual text comparison. This preprocessing step allows rapid lookup and comparison during inference, significantly reducing computation time while maintaining precise similarity measurements through the use of pre-trained semantic representations
Solution Approach 2:
The patent uses word embedding vectors as copies of semantic meaning, allowing efficient computational operations on these vector representations rather than manipulating the original text. These vector copies enable fast mathematical operations that preserve semantic relationships while reducing time complexity
3Productivity
If Word Mover's Similarity measures are used, then productivity is improved, but device complexity increases
Solution Approach 1:
The patent replaces traditional mechanical text comparison operations with vector-space mathematical operations. Instead of comparing text strings character-by-character or word-by-word using complex algorithms, it substitutes these with efficient vector arithmetic operations on embedding representations, achieving faster computation with manageable algorithmic complexity
Data Source
AI summary
Various embodiments of the present invention provide methods, apparatus, systems, computing devices, computing entities, and/or the like for performing text similarity determination. Certain embodiments of the present invention utilize systems, methods, and computer program products that perform text similarity determination by using at least one of Word Mover's Similarity measures, Relaxed Word Mover's Similarity measures, and Related Relaxed Word Mover's Similarity measures.