Database Query Response Using Semantic Vector Similarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing database systems are unable to effectively respond to queries by grouping or aggregating data based on semantic similarity rather than exact value matching, limiting their ability to return relevant results in hierarchical or relational data structures.
Innovation Solution
A method that converts attribute-value pairs in database records into tokens, which are then vectorized using an embedding model to determine semantically similar rows based on similarity measures like cosine similarity, allowing for query responses that utilize rows with values semantically similar to the target row.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If exact value matching is used for database queries, then query precision is improved, but query relevance and semantic understanding deteriorate
Solution Approach 1:
The patent transforms the query matching parameter from exact value equality to semantic similarity based on vector representations. Instead of comparing discrete attribute values for exact matches, the system converts values into vector embeddings and computes similarity metrics (e.g., cosine similarity), enabling queries to return semantically relevant results rather than only exact matches.
Solution Approach 2:
The patent introduces vector embeddings as an intermediary representation between the original attribute values and the similarity computation. The embedding model transforms discrete values into continuous vector spaces where semantic relationships are preserved, allowing the system to measure semantic similarity through vector distance metrics rather than direct value comparison.
2Adaptability or versatility
If semantic similarity computation is implemented, then query versatility and relevance are improved, but computational complexity and processing time worsen
Solution Approach 1:
The patent pre-computes and stores vector embeddings for all attribute values in the database before query execution. This preliminary action transforms the complex semantic similarity computation into a simpler vector comparison task during query processing, reducing the computational burden at query time while maintaining semantic versatility.
Solution Approach 2:
The patent replaces the mechanical exact-matching mechanism with a mathematical vector similarity computation system. Instead of using discrete equality checks, the system employs continuous vector space mathematics and similarity metrics, substituting a computationally intensive but semantically rich approach for the simple but semantically limited exact matching.
3Adaptability or versatility
If vectorization and embedding models are used, then semantic understanding is improved, but system complexity and resource requirements worsen
Solution Approach 1:
The patent employs a universal embedding model that can represent multiple types of attribute values (text, categories, identifiers) within a single vector space. This multi-functional approach allows the same vectorization infrastructure to handle diverse data types, reducing overall system complexity compared to implementing separate matching mechanisms for each data type.
Data Source
AI summary
From a first attribute-value pair in a record, new data is created including a first token. Using a first model and using a processor and a memory, each token is vectorized into new data including a corresponding vector. From the record, a target row is selected, wherein a target attribute-value pair in the target row includes a value for which a semantic similarity computation is to be performed. Using a similarity measure, a set of most similar rows to the target row is determined, wherein each row in the set of most similar rows to the target row has a corresponding similarity measure above a threshold similarity measure and wherein each row in the set of most similar rows includes the target attribute. The set of most similar rows is used to compute a response to a database query.