Blog & News

Looking forward to INLG 2019 in Tokyo

By Cammy Moore | October 25, 2019

Arria’s Chief Scientist & University of Aberdeen Professor Ehud Reiter and Applied Research Scientist Craig Thomson will be representing Arria at the 12th International Conference on Natural Language Generation in Tokyo next week, October 29 – November 1, 2019. Arria NLG is proud to provide thought leadership at this event, the largest annual gathering of NLG expertise in the world.

The conference – sponsored by Google, Facebook, and other global companies – includes invited lectures, presentations, and workshops.

Ehud’s involvement in the conference includes a workshop called Interactive Natural Language Technology for Explainable Artificial Intelligence. Explainable AI (or XAI) is the term for artificial intelligence systems that can explain their AI reasoning to human users. Ehud’s talk is based on a paper he authored, which highlights four specific challenges that are important in generating good explanations of AI reasoning: evaluation, vague language, narrative, and data quality.

Craig is taking part in a workshop on discourse structure in neural NLG (NNLG). NNLG systems learn to generate text from data. Such systems are often end-to-end or “one step” in that they learn and generate natural language sentence realizations in one step. Craig explained that one of the more exciting things happening this year at the conference is that many papers (as well as the NNLG workshop) are actively discussing moving away from end-to-end systems.

In addition to the workshop, Craig will be presenting findings from his current doctoral work, which looks at combining machine learning with rules-based NLG systems, such as those offered by Arria.

Ehud said he is looking forward to catching up with people he has known for years, seeing what they’re doing, as well as meeting new NLG researchers. He went on to say “This year we expect to see progress on neural-network and deep-learning techniques in NLG, particularly with the reduction of hallucination, which is when black-box systems inexplicably output incorrect content. We also look forward to reviewing a lot of exciting work on better techniques for evaluating NLG systems.”

Craig is excited about his first visit to Tokyo, and sees the conference as a valuable opportunity to meet and talk with other researchers. “Being able to speak with others about their work, as well as your own, can help you see things from different perspectives. The conference is a great environment for that.”

A little more about Craig: Arria is lucky to have him with us on a doctoral studentship from EPSRC. (EPSRC is short for Engineering and Physical Sciences Research Council.) He is working on figuring out how recent academic advances can be incorporated into Arria’s software, developing and testing prototypes, thus ensuring that Arria continues to broaden its competitive advantage as the most advanced NLG platform.