I’ve seen a lot of discussion recently about how the academic computer science, artificial intelligence, and computational linguistics (NLP) communities are biased against women and people from racial minorities.  To this, I would add that we are also biased against people with disabilities.  These biases are shameful, and need to change.   In the 30 years I have been attending conferences, for example, although I have seen a number of keynotes from women (but nowhere near 50%), I have only seen *one* keynote at a major event by a black person (and this was at a professional event for AI developers, not at an academic conference).   I’ve never seen a keynote from someone who was obviously disabled.  Which is crazy!  Especially considering that the best undergrad I taught last year was a woman, and the best MSc student I taught last year was black.

Given this background, I would like to say that I am hugely impressed by the diversity (in all of the above senses) of Maria Keet’s research group at Capetown.   Maria’s group focuses on NLG in Bantu languages, which is a very important but under-researched topic (the NLP community suffers from inadequate diversity in the languages it focuses on as well as in its researchers).    They focus not on corpora but on computational modelling of the underlying linguistics, which I must say appeals to me!  This is largely because of the limited availability of corpora in Bantu languages.   For the larger Bantu languages, such as isiZulu and isiXhosa, its possible to get generic corpora to train language models, but its not possible to get the kind of corpora we need for NLG.  I remember a discussion about generating weather forecasts with one of Maria’s students, Zola Mahlaza; I told him to build up a corpus, and he told me that it was almost impossible to get hold of isiXhosa forecasts which were not just translations of English forecasts.   For the smaller Bantu languages, such as Runyankore (which is spoken by millions of people), even getting good generic corpora is a challenge.

Anyways, going back to diversity, I’ve met several of Maria’s students at various events (and also twice served as a thesis examiner for Maria’s students), and they are a very diverse bunch from all of the above perspectives (as well as being very competent and switched-on).   Perhaps the most impressive from a diversity perspective is Joan Byamugisha, who is not only black and female (and has major problems attending conferences because of her citizenship), but is also blind.    Joan should get her PhD soon, which shows that blindness cannot stop a good researcher!

So, my congraulations to Maria for building up such a diverse research group, and showing the rest of us what is possible!

One final thought is that I wonder if one of the best ways to encourage diversity (in languages as well as people) is to encourage and support researchers and students who are outwith the “known” research centres in Europe, North America, and East Asia. Capetown is the best university in Africa, but very few of the researchers I interact with seem to be aware of it.  If we do a better job of encouraging colleagues at diverse universities around the world, this will make our field more diverse in other ways as well.