Cascaded Language Interface for API Access

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing natural language interfaces for computing systems are limited in their ability to understand and process unstructured natural language inputs, particularly in accessing and invoking computing applications and resources without requiring specific syntax knowledge.

Innovation Solution

A cascaded language interface system that uses zero-shot classification, sentence embedding, and extractive question answering models to convert unstructured natural language inputs into structured API calls, allowing users to interact with computing applications and resources without needing to learn API syntax.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If a natural language interface is implemented to allow users to access APIs without learning syntax, then ease of operation is improved, but the system's ability to accurately understand and convert unstructured language into structured API calls deteriorates

Engineering Contradiction:

Improveease of API accessVSAvoidlanguage understanding accuracy

Core Design Contradiction:

Ease of operationVSMeasurement precision

Solution Approach 1:

The patent introduces an intermediary natural language interface layer between the user and the API system. This interface includes components such as a syntax analyzer, semantic analyzer, and API call generator that mediate the conversion of unstructured natural language into structured API calls, thereby maintaining both ease of operation and understanding accuracy

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces the traditional mechanical syntax-based API calling mechanism with a natural language processing system. Instead of requiring users to follow rigid syntax rules, the system uses linguistic analysis and semantic understanding to interpret user intent and generate appropriate API calls, substituting structural compliance with intelligent interpretation

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Ease of operation

If a natural language interface is implemented to eliminate syntax requirements, then ease of operation is improved, but reliability of API call generation deteriorates

Engineering Contradiction:

Improveease of API accessVSAvoidAPI call generation reliability

Core Design Contradiction:

Ease of operationVSReliability

Solution Approach 1:

The patent implements feedback mechanisms where the system analyzes the generated API calls against the original natural language input and validates them against the target system's expected format. This feedback loop ensures that the conversion from natural language to structured calls maintains reliability by verifying both semantic fidelity and syntactic correctness

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent performs preliminary analysis of the natural language input using syntax and semantic analyzers before generating API calls. This preliminary processing identifies potential issues early in the conversion process, allowing the system to correct ambiguities and ensure reliable call generation before the actual API invocation

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS20250103407A1Natural language interface for an application programming interface via cascading zero-shot language models

Publication Date: 2025.03.27 DELL PROD LP

AI summary

Generating a call or request from unstructured input. Natural language, which is unstructured, is received and processed by a language interface engine. The language interface engine executes a series of models in a cascaded manner. One model selects an application programming interface from a set of target application programming interfaces. Another model then selects a particular method of the selected application programming interface. Another model then uses values from the unstructured input as arguments to include in the call. The call is then generated and executed.