Published on Monday 27 November 2023 by QUT Digital Observatory
Last week, the DO team had the wonderful opportunity to present at Research Bazaar (ResBaz) Queensland 2023. A popular conference in the research community, ResBaz holds an annual three-day event featuring keynote speeches, digital research training workshops, and roundtables, all aiming to upskill and connect researchers from diverse backgrounds.
As in previous ResBazzes, the DO team did a number of workshops and talks related to our areas of expertise. Of note this year were two workshops on AI for research, and one showcasing our new data sources.
Having done a lot of investigation and experimentation with Large Language Models (LLMs), Robert Fleet and Mat Bettinson did two talks on the capabilities and potential of these powerful tools in aiding research. Specifically, we discussed the characteristics and properties of LLMs, their strengths and weaknesses, and based on those, how to leverage the power of AI to augment research.
Next, we gave a demonstration of some actual AI-incorporated workflows to analyse a prepared dataset. The scripts, written in Python, dealt mostly with summarisation tasks, from extracting the different viewpoints to identifying the sentiments related to a topic. Attendees got to see how OpenAI's powerful GPT models can be used to assist with exploring large quantities of text data.
In our final session at ResBaz 2023, we introduced exciting new tools that the DO had recently developed - Youte+ and NewsTalk. Building on top of youte, a command-line tool to collect and tidy YouTube metadata, Youte+ is a web interface that makes the process easier for researchers without a technical background. The tool - which is still in the alpha stage, is publicly available and we're keen to get feedback from users trying it out.
NewsTalk is a collection of reader commentary on online Australian news websites. Particularly, on-site and/or Reddit comments linking to Australian news articles are harvested and made ready for analysis. We introduced other features of NewsTalk, including those that enable users to identify similar stories and download comments.
Lastly, we showed a demo of text analytics on the data collected from Youte+ using DSTaR, a R package developed by our Robert Fleet to perform discriminatory text analysis on short texts.
If you have any questions regarding any of our workshops, or simply want to know more, feel free to reach out to the DO via digitalobservatory@qut.edu.au.
