'Language-Based Pipelines to Deliver Value'
9am - 1:45pm, 15th August 2024
Venue : KineticIT, Level 11, 30 The Esplanade, Perth
Data.Conversations is an invitation only workshop series by the UWA Data Institute aimed at people at manager level and above whose teams are using or considering use of AI and who have input into strategic business decisions relating to data, data storage and analytics.
We invite attendees who are involved in assessing potential opportunities for the introduction of AI to improve business process. In this fast-moving space, you want to understand more about potential pipelines, capabilities, data, and time required, and hear about others’ experiences. You are curious (but cautious) and value the opportunity to ask questions. This half day workshop is about having these conversations.
The workshop format is based on a series of short talks, panels, and small group discussions with participants able to interact with the speakers and each other.
This inaugural Data.Conversations workshop is titled ‘Language-Based Pipelines to Deliver Value’ and focuses on Natural Language Processing, Generative AI, and Language Models. Some have estimated that as much of 80% of corporate data is stored as unstructured text. How do you get value from this?
See bottom of page for Workshop Contacts
Leading image generated by ChatGPT4o
PROGRAM
Registration from 8:30AM
9:00am – 9:05am
Welcome Address
Professor Michael Small
Director UWA Data Institute
CSIRO-UWA Chair of Complex Systems
9:05am – 9:15am
Introduction and the Practical Benefits of AI in Processing Unstructured Texts
Professor Melinda Hodkiewicz
Session 1: 9:15AM - 10: 45AM
9:15am – 9:35am
Tyler Bikaun
Garbage In, Garbage Out: Navigating Technical Text Quality – Current Challenges and Future Directions
The effectiveness of artificial intelligence (AI) and natural language processing (NLP) technologies heavily relies on the quality of textual input. This challenge is especially pronounced in technical domains such as industrial mining and maintenance. In this presentation, Tyler will examine how unstructured textual content affects the performance of AI models and their applications. Through real-world examples he will illustrate the business implications of compromised text quality, including reduced efficiency and accuracy in AI applications. Tyler will also discuss contemporary solutions, such as narrow and generative AI techniques, evaluating their potential and inherent limitations.
9:35am – 9:55am
Dr. Michael Stewart
Automating technical text classification using AI/NLP (Natural Language Processing)
Unstructured technical texts, such as maintenance work orders and safety records, capture a wealth of valuable historical business information. Michael will demonstrate use of NLP tools in automating the classification of technical texts. Failure Mode Classification on Maintenance Work Orders is used to showcase how AI can be used to quickly perform a task that takes engineers ages and they do inconsistently. He will discuss the value of investing in high quality in-company human-annotated data in training these models to perform effectively. There is considerable opportunity for managers to use NLP improve the quality of many classification tasks, especially important when these classifications drive downstream business improvement programs and decisions.
9:55am – 10:40am
Two hosted table-based conversations
(20 mins each)
Morning Tea 10:40AM - 11:10AM
Session 2: 11:10AM - 12:40PM
11:10am – 11:30am
Caitlin Woods
Linked Data: The Foundations of Enterprise AI
In this presentation, Caitlin will explore the role of linked data and semantic technologies in the future of enterprise AI. We will discuss how linked data supports data interoperability, FAIR data principles, and data quality. Additionally, we provide practical examples of how linked data can be developed, maintained, and governed in a real-world organisation. Why as an operational or technical leader should this be of interest? Come and learn why Linked Data is emerging as a foundational practice for information exchange and interoperability in asset intensive organisations in Europe and why Linked Data schemas needs to be fit-for-purpose for the business context and grounded in the needs of engineers, managers and
decision makers.
11:30am – 11:50am
Professor Melinda Hodkiewicz
Unpacking language AI risks and opportunities for operations and technical leaders
In preparing for this workshop Melinda interviewed 25 technical and operations leaders in Perth. In this talk she will summarise the questions raised and try to address some topics such as “What are the practical benefits of AI in processing unstructured texts and in technical records management?”, “Give us some examples of use cases and emerging products?”, “What are the cost and ROI of implementing AI solutions?” and “What skills and training will our teams need to leverage these AI tools?”
11:50am – 12:35pm
Two hosted table-based conversations
(20 mins each)
12:35-12:45pm Wrap up - Professor Michael Small and A/Professor Wei Liu
12:45pm - 1:45pm Lunch & Networking
Workshop Queries
Content: Professor Melinda Hodkiewicz, UWA School of Mechanical & Chemical Engineering, melinda.hodkiewicz@uwa.edu.au
Logistics: Julianne Sparke, UWA Data Institute, uwadatainstitute@uwa.edu.au