Contextual Keyword Matching with Phrase-Localized Attention
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional keyword matching systems produce a high number of false positives due to the inability to understand context, leading to irrelevant results, especially in business applications where words like 'apple' can refer to both a company and a fruit.
Innovation Solution
Implementing an encoder with phrase-localized attention layers and scaled dot-product attention layers to learn the local structure and relationships between keywords, using keyword-level positional encoding to ensure permutation invariance, thereby improving contextual keyword matching accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional string matching is used for keyword matching, then the matching process is simple and fast, but it produces a high number of false positives and cannot understand context
Solution Approach 1:
The patent replaces traditional mechanical string matching algorithms with an AI-based encoder system that uses attention mechanisms and embedding vectors to perform contextual keyword matching, thereby improving accuracy while accepting increased system complexity
Solution Approach 2:
The system transforms discrete keywords into continuous embedding vectors and uses attention weights as dynamic parameters to capture contextual relationships, fundamentally changing the matching approach from exact string comparison to semantic similarity assessment
2Reliability
If contextual awareness is added to keyword matching, then false positives are reduced, but the computational complexity and processing time increase
Solution Approach 1:
The patent pre-trains the encoder model on large datasets and pre-computes embedding vectors for keywords, so that during actual matching operations, the system can quickly retrieve and compare pre-processed representations rather than performing complex computations in real-time
Solution Approach 2:
The attention mechanism divides the keyword array into individual elements and computes attention weights for each, allowing the system to focus computational resources on the most relevant keywords while still maintaining contextual awareness across the entire array
Data Source
AI summary
State-of-the-art keyword matching may result in a high number of false positives, since computers are unable to understand context in the same manner as humans. Accordingly, artificial intelligence for contextual keyword matching is disclosed. In particular, the artificial intelligence may comprise an encoder that comprises one or more phrase-localized attention layers, with keyword-level positional encoding to ensure permutation invariance. Each phrase-localized attention layer may comprise a multi-head phrase-localized attention network for each keyword in an input keyword array. The encoder may also comprise one or more scaled dot-product attention layers, subsequent to the phrase-localized attention layer(s). The phrase-localized attention layer(s) enable the encoder to learn the local structure of the keywords, while the scaled dot-product attention layers enable the encoder to learn the relationships between the keywords. This improves the accuracy of the contextual keyword matching, which may, in turn, improve the accuracy of downstream functions.