Semantic Database Queries via Word Vector Representation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current database systems require users to have knowledge of the database schema to extract information, making it difficult to navigate and query large, complex databases effectively, especially when dealing with multiple tables and vast amounts of data.
Innovation Solution
The integration of word vector representation technology into relational databases, allowing for semantic-based queries that encode relationships between database entities using vectors, enabling users to query and navigate databases without detailed schema knowledge by representing text entities and their relationships numerically.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If users query databases using traditional SQL with schema knowledge, then precise information extraction is achieved, but user operation complexity increases and accessibility decreases
Solution Approach 1:
The patent introduces word vectors as an intermediary layer between user queries and database schema. Instead of requiring users to directly interact with complex database schemas, the system converts both user intent and database entities into vector representations, where semantic similarity guides the query process. This intermediary vector space simplifies user interaction while maintaining precise information extraction capabilities.
Solution Approach 2:
The patent transforms the query interface from syntactic SQL parameters to semantic vector parameters. By representing database entities, user queries, and relationships as vectors in a continuous semantic space, the system allows users to query based on semantic meaning rather than requiring knowledge of specific schema parameters, table structures, or SQL syntax.
2Quantity of substance
If databases store vast amounts of data across multiple tables, then information capacity increases, but navigation and query efficiency decrease
Solution Approach 1:
The patent adds a semantic vector dimension to traditional database navigation. Instead of navigating solely through tabular relationships and joins, the system projects database entities into a vector space where semantic relationships are preserved. This allows efficient navigation by querying vectors for semantically similar entities without traversing multiple table joins, significantly improving query efficiency across large databases.
Solution Approach 2:
The patent replaces the mechanical system of SQL joins and schema traversal with a semantic vector-based retrieval mechanism. Instead of mechanically navigating through database schemas and joining tables based on foreign keys, the system uses vector similarity calculations to directly retrieve relevant entities, substituting complex mechanical navigation with efficient mathematical operations in vector space.
3Measurement precision
If databases use traditional syntactic query operations, then exact matching is achieved, but semantic relationship discovery is limited
Solution Approach 1:
The patent merges syntactic matching and semantic relationship detection into a unified vector-based query system. By representing both exact matches and semantic relationships in the same vector space, the system can simultaneously achieve precise matching when needed while also discovering semantic relationships through vector similarity, combining the strengths of both approaches.
Solution Approach 2:
The patent introduces dynamic query capabilities where the matching behavior adapts based on the vector similarity threshold. Users can adjust the threshold to balance between exact matching (higher threshold) and semantic relationship discovery (lower threshold), making the system dynamic and adaptable to different query needs rather than being fixed to either syntactic or semantic matching.
Data Source
AI summary
A method, apparatus, and computer program product are provided for generating a set of token sequences for at least a portion of a database, wherein each token in a sequence represents a respective database entity of the database; assigning, for each token in the set of token sequences, at least one corresponding vector from a set of vectors of a same dimension, wherein the at least one corresponding vector encodes relationships between the database entity of a token and other database entities of other tokens of the set of token sequences; and extracting, using a query language, information from the database based at least in part on the relationships encoded by the assigned vectors.