Expert Group Meeting Big Data & AI Technologies

By Kurt Stockinger, ZHAW and Thierry Bücheler, Oracle

On November 9, 2022 the Expert Group meeting was held directly before the General Assembly of the Data Innovation Alliance, which gave the group an extra boost.

The first talk was given by Thierry Bücheler from Oracle titled “De-buzz AI”. The talk described various AI uses cases as part of customer projects with Oracle and introduced some of Oracle’s AI technology solutions in the context of current Data Science (DS) projects across industries. The main focus was on “real-world” challenges and soft factors: Thierry discussed the importance of provisioning, managing, and sharing persistent DS notebooks, how to combine data silos, the collaboration between data scientists/ML experts and domain specialists who are not necessarily IT-savvy, import/export of code (e.g., Python) to GUI interfaces and visualizations for non-coders, lifecycle management and efficient re-use of models, automatic data pre-processing, the connection to business workflows and -goals, as well as the connection of open source software and libraries with “enterprise-hard” (secure, available) proprietary platforms for scaling and distributed collaboration. He also did a deep dive on anomaly detection based on modified MSET2 and which “advanced” ML tools you can get out-of-the-box nowadays without much coding.

The second talk was given by Kurt Stockinger from Zurich University of Applied Sciences titled “Talking to Data: Building Natural Language Interfaces for Databases”. First, the challenge of querying large databases was introduced, where end users need to be proficient in database languages such as SQL and SPARQL. Afterwards, Kurt introduced two different approaches for automatically translating natural language questions to either SQL or SPARQL. The first approach, called Bio-SODA, is based on a pattern matching algorithm that does not use machine learning. The second approach, called ValueNet, is based on complex neural network architectures leveraging state-of-the-art transformers. The algorithms have been applied to query several databases from the areas of astrophysics, bioinformatics but also from industry. Finally, Kurt also showed how ValueNet has been used to a world cup database with information on games dating back to the first world cup in Uruguay in 1930. The demo has been built in collaboration with the ZHAW Institute of Information Technology and the ZHAW Centre for Artificial Intelligence.

More information about using AI technology to access data can be found at the web site of the European Union Project called INODE – Intelligent Open Data Exploration.