AI Search Ranking via Semantic Vector Similarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing search engines rely on text similarity for sorting documents, which results in low efficiency and inaccuracy due to slow matching calculations and inability to determine ordinal relations between segmented words.
Innovation Solution
A search method and apparatus using artificial intelligence that converts query and document sentences into semantic vectors through neural networks, calculating similarity between these vectors to rank documents accurately.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If text similarity calculation is used for sorting documents, then the sorting can be performed based on simple text matching, but the sorting efficiency is low and the sorting accuracy is not enough
Solution Approach 1:
The patent replaces traditional text similarity calculation (mechanical string matching) with neural network-based semantic vector representation. The neural network model transforms text into semantic vectors that capture meaning rather than just character similarity, thereby improving sorting accuracy while maintaining computational efficiency through optimized vector operations.
Solution Approach 2:
The patent changes the fundamental parameter used for text comparison from character-level similarity to semantic vector similarity. By representing texts as vectors in a high-dimensional semantic space and calculating cosine similarity or other distance metrics between vectors, the system achieves both higher accuracy in understanding text meaning and improved efficiency through vectorized computations.
2Ease of manufacture
If traditional text matching is used, then the implementation is simple, but the ordinal relation between segmented words cannot be determined
Solution Approach 1:
The patent replaces simple text matching with neural network-based semantic vector representation, which inherently captures word order and contextual relationships. The sequential processing of words through the neural network preserves ordinal information in the resulting semantic vectors, enabling accurate determination of word relationships without complex additional processing.
Data Source
AI summary
Embodiments of the present disclosure disclose a search method and apparatus based on artificial intelligence. A specific implementation of the method comprises: acquiring at least one candidate document related to a query sentence; determining a query word vector sequence corresponding to a segmented word sequence of the query sentence, and determining a candidate document word vector sequence corresponding to a segmented word sequence of each candidate document in the at least one candidate document; performing a similarity calculation for each candidate document in the at least one candidate document; selecting, in a descending order of similarities between the candidate document and the query sentence, a preset number of candidate documents from the at least one candidate document as a search result.