Machine Learning Research Group

The mission of the Machine Learning Research Group is to scale Machine Learning across Oracle by researching and developing ML-based solutions that improve Oracle's products and services.

Project Overview

Mission

The mission of the Machine Learning Research Group (MLRG) is to scale Machine Learning (ML) across Oracle. We do this by undertaking fundamental ML research. MLRG members' research interests cover a wide range of ML work. We collaborate with product groups and business units world-wide in order to develop ML-based solutions that will improve their products and services.

We build proof-of-concept systems that demonstrate the viability of an ML approach to solving particular business problems. These proof-of-concept systems and methodologies are transferred to the product group.

We help Oracle's product groups understand how to operationalize Machine Learning solutions in order to get the maximum benefit from them.

We acquire and build tools for our use and the use of the ML community in Oracle. We are responsible for the Tribuo Machine Learning package.

We help build an ML community at Oracle. The MLRG hosts an internal ML summit each year that brings together hundreds of ML practitioners from across Oracle.

Collaborations

The MLRG collaborates with a number of groups inside Oracle.

Oracle Sales

We have been working with Oracle's Sales team to understand how we can use ML to optimize Oracle's sales process. We have built lead scoring algorithms that identify the leads that are most likely to convert and are working on models that help us understand our customers future product needs.

Fusion Financials

We have been working with the Fusion Financials team to understand how Natural Language Processing techniques can be used to automatically extract useful information from invoices and receipts.

Oracle Digital Assistant

We have been working with the Oracle Digital Assistant team on a variety of Natural Language Processing projects, including investigating scalable, fault-tolerant, multi-lingual contextual embeddings.

Research Interests

The MLRG's research interests span a wide range of topics in Machine Learning and Natural Language Processing. We are pursuing research projects in aspects of Machine Learning devoted to fairness and privacy. In particular, we are trying to understand how these aspects of ML apply to the business problems that Oracle's product groups are focused on.

In the realm of NLP, we have built systems for model-based sentiment analysis, named entity recognition, entity linking and coreference resolution and product attribute extraction. The MLRG is responsible for one of the most scalable topic modeling algorithms that can train a model on billions of documents in hours on a cluster of computers. We are currently working on approaches to building large-scale, multilingual contextual embedding models that are resistant to many types of errors.

We are working on standard image recognition models and trying to understand how those models might help in training image recognition models for cervical cancer detection.