Feature Store Queries via Natural Language and Zero-Shot LLMs

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Accessing and querying data in feature stores requires technical proficiency, posing a barrier for users with limited expertise, and traditional methods involve complex query languages, hindering accessibility and usability.

Innovation Solution

A natural language processing approach using a large language model (LLM) translates user queries into optimized programming language queries for feature stores, leveraging schema information and zero-shot learning to generate accurate and efficient queries.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional query languages and complex programming interfaces are used to access feature store data, then query accuracy and system control are improved, but user accessibility and ease of operation deteriorate

Engineering Contradiction:

Improvequery accuracyVSAvoiduser accessibility

Core Design Contradiction:

Measurement precisionVSEase of operation

Solution Approach 1:

The patent introduces a natural language processing intermediary that mediates between the user and the feature store system. Users submit queries in natural language, which the NLP system translates into structured programming language queries that the feature store can execute. This intermediary layer maintains query accuracy while eliminating the need for users to learn complex query languages or programming interfaces.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces the mechanical system of traditional query languages and programming interfaces with a natural language processing system. Instead of requiring users to manually construct queries using specialized syntax and protocols, the system uses NLP to automatically interpret and translate natural language requests into executable queries, significantly reducing the operational barrier.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Productivity

If specialized query languages and programming knowledge are required to access feature stores, then system control and query precision are improved, but productivity and collaboration across teams deteriorate

Engineering Contradiction:

Improvecollaboration efficiencyVSAvoidsystem complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent makes the feature store system universal by enabling multiple user groups with varying technical backgrounds to access the same data repository using a common natural language interface. The system maintains its complex functionality and control mechanisms while providing a simplified, universal access layer that allows data scientists, analysts, and business users to collaborate efficiently without requiring specialized programming knowledge.

Inventive Principle:

Principle #6Universality (Multi-functionality)

3Measurement precision

If complex programming language queries are used to retrieve features, then data retrieval accuracy is improved, but query time and learning curve increase

Engineering Contradiction:

Improvedata retrieval accuracyVSAvoidquery preparation time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs preliminary action by pre-processing and understanding the natural language query to generate the optimized programming language query automatically. The NLP system analyzes the user's intent, identifies relevant features and parameters, and constructs the appropriate query structure before execution, eliminating the time users would otherwise spend learning and writing complex queries while maintaining retrieval accuracy.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS12373425B1Natural language query generation for feature stores using zero shot learning

Publication Date: 2025.07.29 ORACLE INT CORP

AI summary

The present disclosure pertains to natural language techniques for querying data stored in feature stores using zero shot learning. In a particular aspect, a computer-implemented method includes receiving a natural language query for retrieving features from a feature store, generating an input prompt by appending a script to the natural language query, and then using a large language model to determine tables or databases from the feature store that are relevant to the natural language query, retrieve metadata for the tables or databases from the feature store, determine feature groups comprising features relevant to the natural language query, and generate a programming language query based on the input prompt, the metadata, and the groups. A list of features within the feature groups that are accessible within the feature store may then be retrieved by executing the programming language query on the feature store.