Invention Title:

A CHATBOT FOR DEFINING A MACHINE LEARNING (ML) SOLUTION

Publication number:

US20250077915

Publication date:

2025-03-06

Section:

Physics

Class:

G06N5/04

Inventors:

Alberto Polleri London, Great Britain (UK)

Marc Michiel Bron London, Great Britain (UK)

Dan David Golding London, Great Britain (UK)

Alexander Ioannides London, Great Britain (UK)

Maria del Rosario Mestre London, Great Britain (UK)

Hugo Alexandre Pereira Monteiro London, Great Britain (UK)

Oleg Gennadievich Shevelev London, Great Britain (UK)

Larissa Cristina Dos Santos Romualdo Suzuki Wokingham, Great Britain (UK)

Xiaoxue Zhao London, Great Britain (UK)

Matthew Charles Rowe Bletchley, Great Britain (UK)

Sergio Lopez London, Great Britain (UK)

Assignee:

ORACLE INTERNATIONAL CORPORATION Redwood Shores, CA, United States

Applicant:

Oracle International Corporation Redwood Shores, CA, United States

Smart overview of the Invention

The patent application introduces a system and method for an intelligent assistant, specifically a chatbot, designed to help users create machine learning systems. This tool simplifies the process for users who may not have expertise in software development or machine learning. By using natural language inputs, the chatbot translates user interactions into structured machine learning solutions, enabling users to develop and deploy models without needing to be data scientists.

Background

Machine learning has diverse applications but often requires significant programming knowledge to build custom models. Existing tools lack interoperability due to varying technologies and languages. Moreover, these tools do not efficiently store model components for reuse. The reconciliation of data from varied sources is cumbersome and existing tools focus on maximizing accuracy rather than adapting pipelines in real-time based on data metric changes.

Key Features

The proposed machine learning platform generates a library of components to facilitate model creation without requiring in-depth knowledge of infrastructure or coding. It analyzes data and user-defined criteria to select library components and APIs for building applications. The platform supports model training, testing, and compilation into executable code, while also allowing users to create profiles for personalized recommendations.

Chatbot Functionality

The chatbot offers an intuitive interface for generating machine learning applications. It guides users in locating data, selecting solutions, displaying optimal options, and recommending deployment environments. The system includes a self-adjusting feature that maps client data schemas to standard classifications, making solutions adaptable across different clients.

Advanced Techniques

The system uses data ontologies for precise service searches and pipeline composition with minimal human intervention. It can generate new ontologies if needed. An adaptive pipelining service incorporates new models into applications, testing them offline against ground truth data. Successful models are automatically promoted to production, ensuring ongoing optimization of the machine learning application.