Professor Richard Kittredge died in early April. Richard was a pioneer in applied NLG, amongst other things. He will be missed! I won’t try to summarise Richard’s career as a whole, or his personal life, but I thought I would write a few words about Richard’s impact on me and my career.
I first met Richard around 1990, when I was finishing up my PhD. The NLG and NLP worlds were much smaller in 1990 than they are today, and at the time Richard was perhaps best known for his pioneering work on the TAUM METEO system for translating Canadian weather forecasts from English to French. METEO was one of first operationally deployed MT systems. I remember Richard explaining to me that the system ran automatically, with humans only checking the small number of translations which the system was not confident in.
The METEO experience led Richard to investigate sublanguages, that is limited and non-standard variants of English which are used in specialised domains such as weather forecasts, recipes, and maintenance manuals. Richard was a leader in sublanguage research.
In the late 1980s, while I was doing my PhD, Richard started investigating whether it was possible to use NLG techniques to generate weather forecasts from the output of a numerical weather model. Which at the time was a crazy idea! Richard and his wife Tanya Korelsky set up a company, CoGenTex, in 1990, which started working on an operational deployable weather forecast generator (not just a research prototype). In 1992, CoGenTex’s FoG system went live at the Canadian weather service. This was a milestone in applied NLG, the first NLG system which was operationally used to produce texts for real-world users! Indeed, even from the perspective of 2019, FoG had some impressive features, such as the ability to generate forecasts in both English and French.
I finished my PhD in 1990, and followed Richard and CoGenTex with great interest. My PhD had been fairly theoretical (basically about models for generating noun phrases which described objects), but I had always been very interested and keen on applied NLG as well as theoretical issues. So I was very impressed and happy when FoG went live, even though I didnt have anything to do with the system; it felt like a huge achievement for the NLG community as well as for Richard and his team.
In fact, I was so impressed that I accepted a job at CoGentex in 1993! I stayed at CoGenTex for two years before resuming my academic career by accepting a faculty position at Aberdeen. While at CoGenTex I worked on several applied NLG projects (most of which were confidential, one exception was ModelExplainer), and also had many fascinating conversations with Richard, other CoGenTex staff, partners, and clients about where NLG technology could best be used in real-world contexts. These experiences have shaped much of my subsequent work on applied NLG.
Richard was a strong proponent of Meaning-Text Theory, which separates the NL process into several stages, such as Deep Sytntax, Surface Syntax, Deep Morphology, and so forth. Indeed, CoGenTex developed its own realiser, RealPro, based on MTT. I guess I personally was never 100% convinced by MTT (some nice ideas but also some problems), but certainly working with Richard on MTT broadened by horizons linguistically, since MTT was very different from the more “standard” grammar formalisms I had been taught as a student.
As mentioned above, I resumed by academic career in 1995, as a Lecturer at Aberdeen University (where I have been ever since). I must admit that my contacts with Richard rapidly diminished, as I focused on the challenges of lecturing to students, writing my book on building NLG systems, and also my smoking-cessation research project. But I still saw Richard occasionally at conferences and interacted with him by email. He was very interested in Data2Text and Arria, and gave me some good suggestions and insights about commercialising NLG.
So, farewell to a pioneer in applied NLG, who created the world’s first specialist NLG company and delivered the world’s first operationally used NLG system, and also influenced and shaped my career in many ways!