Semantic Data Retrieval Cost Function Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data retrieval methods, such as SQL queries, require users to specify the tables from which data is to be selected, leading to complex queries and inefficient performance due to the need for runtime JOINs.
Innovation Solution
A computer-implemented method that receives a query with semantic information, determines if the query can be serviced using data from candidate data sets, and uses a cost function to select the appropriate data sets and portions thereof to return the requested data without specifying the tables.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If SQL queries require users to specify tables and JOINs, then data retrieval can be performed, but query complexity increases and performance decreases due to runtime JOINs
Solution Approach 1:
The patent introduces an intermediary layer (the system) that translates semantic queries into optimized data retrieval operations. Users submit queries in natural language or semantic format without specifying tables or JOINs, and the system automatically determines the necessary data sources and relationships, eliminating the complexity burden from users while maintaining retrieval capability.
Solution Approach 2:
The system performs preliminary analysis of data relationships and optimal retrieval paths before executing queries. By pre-processing and understanding the data model structure in advance, the system can optimize query execution plans and avoid expensive runtime JOIN operations, thereby improving performance before the actual query execution occurs.
2Ease of operation
If users must identify tables and specify JOINs manually, then precise data retrieval is possible, but the ease of operation deteriorates due to requiring knowledge of the entity-relational model
Solution Approach 1:
The system enables self-service data retrieval where users can query data using semantic information without needing to understand the underlying table structures or relationships. The system automatically interprets the semantic intent, identifies relevant data sources, and retrieves the correct information, making the operation as simple as expressing the data need semantically.
Solution Approach 2:
The system provides a universal query interface that accepts various forms of semantic input (natural language, structured queries, etc.) and automatically adapts to retrieve the required data regardless of the underlying data model complexity. This multi-functional approach allows the same simple interface to handle diverse query types while maintaining accuracy.
3Productivity
If a fully normalized data-warehouse is used, then data organization is efficient, but runtime JOINs are required which slow down query execution
Solution Approach 1:
The system performs preliminary determination of the optimal data retrieval strategy by analyzing the semantic query and pre-identifying the necessary data sources and their relationships. This advance planning allows the system to execute optimized retrieval operations without requiring complex runtime JOIN processing, thereby improving execution speed while maintaining normalized data warehouse benefits.
Data Source
AI summary
A method of retrieving data from at least one data store which stores a plurality of data sets each having a tabular data structure, the method including: receiving a query which includes semantic information and requests data associated with the semantic information from the data store(s); determining from the semantic information whether the query can be serviced using data selected from and/or derived from one or more candidate data set of the plurality of data sets; if multiple candidate data sets can service the query, using a cost function to determine at least one candidate data set of the multiple candidate data sets to service the query, and determine a portion of each of the at least one candidate data set to service the query; and returning a response to the query, the response including data obtained using the portion of each of the at least one candidate data set.