1. What does your role entail as Chief Scientist at Arria NLG?
As Chief Scientist I spend a lot of time developing Arria’s core NLG technology. I have worked as a researcher in NLG since the late 1980s, and was one of the founders of Data2text in 2009. My challenge as Chief Scientist is to translate conceptual ideas about NLG (including lots of great ideas from Arria’s developers) into workable and useful code. There is a huge amount Arria can do in NLG, the biggest challenge is probably prioritizing.I also work with clients, especially on projects in oil and gas and medicine. I have worked on oil and gas NLG projects since 2009, and medical NLG projects since 1995. I think there is enormous potential to use NLG in both of these areas, and I really enjoy working with clients and Arria’s delivery teams to design and build solutions.
2. What excites you about Arria NLG Technology?
NLG humanizes information. There is an enormous amount of information which has the potential to help us, but the reality is that we often end up drowning in information rather than benefiting from it. This is true for professionals such as doctors and engineers, and even more true for ordinary citizens. For example, in many countries people have the right to see their medical record, but a record full of incomprehensible acronyms and numbers is not going to help many people, indeed it may scare people. NLG technology allows us to build systems which summarize and explain information, in understandable language, for different audiences; I believe that it will revolutionize how people interact and use data.
3. You’ve been instrumental in a number of Arria patents, which are you most proud of and found to be the most innovative?
Like most writers, I am most excited by the patents I am currently developing, not the ones that have already been published. That is, novelists are always most excited by the new novel they are in the process of writing, not by a novel they wrote several years ago; the same is true for me as a patent writer.
But with that said, I am very excited by the latest patent we have announced, on the generation of referring expressions. This is an area which I have worked on for many years as an academic; indeed my 1990 PhD was partially about reference (a different kind from the Arria patent). It is very satisfying to see my decades-long interest in reference leading to patents as well as academic publications, and I also believe that reference is something which is essential to high-quality narrative, and which Arria does better than anyone else.
4. What would be your ideal Use Case and why?
I personally am hugely excited by the potential of using NLG to help people manage their own health. By this, I mean apps that take in medical data (medical records and sensor data), and produce easy-to-understand texts that help people live healthier lives (e.g. diet, exercise, smoking), manage chronic conditions (e.g. diabetes, asthma), and comply with treatment regimes (e.g. take medication appropriately). The biggest challenge in healthcare in 2016, at least in the developed world, is to help people manage their health better, and I think NLG can make a tremendous contribution here.
5. Where do you see NLG technology in 5 years?
There are a lot of great ideas floating around in Arria and the wider NLG world, and of course it’s hard to predict which ones will take root and grow. For what its worth, I think that in 2021 we will be much better at producing high-quality narratives, because we’ll be using better models of narrative coherence and structure; these will be motivated at least in part by research on narrative in psychology and literature. We willl be much better at personalizing narratives and texts for different users depending on their task, expertise, personality, etc. Also we’ll be able to automatically derive personalization models from corpora (e.g. we’ll create a personalization model for a user by automatically analyzing his/her emails). Finally, Arria will be much better at producing multimodal interactive documents which integrate narratives and visualizations; this will be driven by a better understanding of the strengths and weaknesses of these modalities for communicating specific types of information and messages, which will come in part from psychological research.
Prof. Ehud Reiter is the Chief Scientist of Arria NLG, and also a Professor of Computing Science at the University of Aberdeen. He has worked on natural language generation for the past 30 years, and has published 125 peer-reviewed academic papers and been awarded 5 patents; he also is the author of a widely used NLG textbook. In his university role he has been involved in several medical projects, and has published papers in British Medical Journal and other leading medical journals.