Conversation with a Language Engineer
I first met Dr. Túndé Adégbọlá in Ìbàdàn around 2002, when he came around to some of our linguistics classes as a guest lecturer. The visits were memorable, not just because he wasn’t a trained linguist but an engineer, but because he had interesting ideas about the direction of language technologies that sounded really fascinating to my undergraduate imagination — words like speech synthesis, automatic speech recognition, language engineering, etc. His passion was hard to miss.
His background was in engineering, but he had become obsessed with language, and looking for ways to combine his abilities with those of trained linguists to create a future that he had already seen elsewhere, and personally envisioned. His dedication began to yield results. In 2004, he launched African Language Technology Initiative (ALT-i), a research and advocacy organisation in Ibàdàn, to formalize his work in that direction. And in August of that year, in collaboration with the Department of Linguistics and African Languages where I was a student, he helped organize the West African Languages Congress (WALC2004) Conference. I was a member of the Local Organising Committee of that conference and handled its website and other duties.
I can say, without a doubt, that my interest in language technology was first sparked during that summer, and I have continued to be inspired by Dr. Adégbọlá’s passion, interest, and involvement in African language technology and engineering. Our paths have crossed many times since then, most recently at the LT4All (Language Technology for All) Conference in Paris, in late 2019, where he had presented a paper. His past experience had been a guide for me while researching the problems of Unicode limitation with Yorùbá writing. I still continue to rely on his direction.
I had known, over the years, that his involvement in our technology ecosystem in Nigeria went deep and wide, but I never quite knew the extent. So I decided to talk to him to put many of his accomplishments, opinions, and interests on the record for others who might be interested. This is a result of that conversation. He continues to work as a teacher, engineer, musician, linguist, and culture activist. His TEDxPortHarcourt talk focused on education as a tool for development. I highly recommend it.
I also hope that you enjoy our conversation.
_____
I’ve known you as one of the pioneers of work in African language technology, because of your work in that direction. Actually, you’re one of the people I have looked forward to for inspiration for the work I do. Do you know since when you’ve been interested in this field?
I started thinking of working on Human Language Technology as soon as I finished my undergraduate studies in Electrical Engineering in 1978. I was privileged to attend the World Conference on Faith, Science and the Future that took place at the Massachusetts Institute of Technology (MIT), Boston, USA in 1979 and I took that opportunity to make inquiries on speech technology. For the first time, I got to use an on-line computer terminal rather than the card punch machines that I had used to write FORTRAN programs for my undergraduate project in the University of Lagos. I combed their libraries and spent quality time in the MIT bookshop and various other book shops in Boston. I therefore came back home with a bounty of valuable materials. Unfortunately, at that time, speech technology demanded substantial financial outlay for the required dedicated equipment which neither my parents nor I could afford, but I kept dreaming. I was subsequently employed as a broadcast engineer and my work in speech processing as an engineer in a radio station helped to sustain my interest in speech technology. Having access to microphones, sound recording equipment, oscilloscopes and signal generators in the station, I was able to sustain my interest by performing simple signal processing experiments. I tried as much as possible to keep abreast of developments in the field mainly through the popular press as I did not have access to the scientific literature. I also bought as many books as I was able to lay my hands on whenever I travelled out of Nigeria.
Can you remember those books? Looking back now, and compared to the current advancements in language technology, I assume they must have been very rudimentary or at best futuristic in nature. Or do you think there were people, even then, that had a clue about how it would work in practice?
Two of the books that fired my passion are Classification Algorithms by Petteri Kaski and Patric R.J. Östergård and Electronic Speech Synthesis by Geofery Bistow. I got the Bistow book in 1985, the year after it was published! What I was engaged in at that time was not rudimentary at all. Most of the physics and mathematics we now use were already well understood. Even though I did not have access to the equipment that can manipulate basic signals into speech in real time, I had a good idea of what was required. It was clear to me that computers would offer the levels of required signal control and so I saved to buy my first personal computer in 1982. I later on took a study leave and went back to the University of Lagos to take a diploma in computer science. I subsequently resigned my appointment with Ogun State Television and set up a computer consultancy in Ibadan. I was very fortunate to be contracted to computerise the stock control of Spectrum Books, a leading publishing company in Nigeria and the Managing Director Mr. Joop Berkhout assisted me in getting all the books that I needed. That gave me a rare opportunity to build a substantial library on speech technology, but I was still not attuned to the academic literature of the subject.
What role did your background in engineering play in this interest? And what have been its limitations?
My background in engineering made a whole lot of difference, so did my foray into computer science but my crass ignorance in linguistics was a major limitation. My main quest at that time was to discover the acoustic correlate of Yorùbá tones. Around 1986, I met Olúbí Johnson and Ṣeéni Ògúnsan, two PhD students in the department of Physics at the University of Ìbàdàn. They were both working on speech technology in their doctoral programs. My library was irresistible and the three of us joined forces trying to make the computer speak and read Yorùbá. Unfortunately, the difficulty in accessing relevant materials and the lack of required equipment made it difficult for us and they both had to change their subjects of enquiry in order to obtain their degrees in reasonable time. Since I was not seeking a degree in speech technology, I just continued my quest until I discovered the subject of linguistics. It was like letting a bird out of a cage, but I had to continue my consultancy work both in engineering and computing to make ends meet. I embedded in the emerging Nigerian motion picture industry (Nollywood) supplying the necessary technologies for motion pictures whilst I continued my research into speech technology on the side.
You did a lot of work with the earlier Mainframe Òpómúléró films. Can you tell me some of your work there, at least for the record?
I worked with many video production houses in Lagos in the early 1990’s, but I was closest to Mainframe. In fact it was Mr. Túndé Kèlání of Maiframe that requested me to come to Lagos to help out with the technology of modern motion picture which was then becoming more and more computer dependent. Having worked as a broadcast engineer from the late 1970’s to mid 1980’s and having acquired specialised knowledge in computing, I had little or no competition in setting up the technology foundations of the then fledgling motion picture industry in Nigeria.
However, Túndé Kèlání’s Mainframe was my base. I helped Mainframe to build the first Hybrid editing machine, a 100 Machine VHS Duplication bay and 3D animation facilities. I was also on hand to solve any technical problems on their film sets and whenever any of their actors disappointed, TK would ask me to act the part. I also composed, performed and did the sound dubbing of a number of Mainframe films such as Ti Olúwa ni Ilẹ̀, Ayo Ni Mo Fẹ́, Ó Le Kú etc. It was just plain fun for me and the exposure that I got from working with Mainframe and other production houses such as Klink Studios, Media International and many others gave me the required profile to be commissioned to design most of the pioneering private TV stations such as Channels, MITV and AIT.
You were a part of setting up those TV channels too?
Oh yes. I designed Africa Independent Television (AIT) from scratch, supplied the equipment, installed and commissioned it. I worked and lived on their premises at Alágbàdo for months between 1996 and 1997. I had also assisted in building Channels Television and MITV before AIT. I also did broadcast engineering work outside Nigeria. I was commissioned by the Open Society Initiative for West Africa (OSIWA) in 2004 to build West Africa Democracy Radio (WADR) in Dakar Senegal. I was also commissioned by UNICEF in 2006 to undertake an assessment of the National Radio and TV of Sao Tome and Principe and later commissioned in 2007 by UNDP, UNICEF and International Alert to build two new radio station in the country in anticipation of the envisaged developments expected from the joint venture oil production they entered into with Nigeria.
Why did you launch ALT-i? Why did you choose 2004 to do it? What did you hope it would achieve, and how are you doing so far in that aspiration?
There is a Yoruba saying; Ẹní wẹjú lẹ̀rù ń bà. It approximates in English as “the clairvoyant holds great responsibility”. I realised very early that the computer was going to become an integral part of everyday life and computer skills were going to become life skills. I also realised that language was going to be the last bridge of the digital divide for many cultures of the world. Guided by the Yorùbá philosophy behind the saying Ẹní wẹjú lẹ̀rù ń bà, I decided to contribute to bridging the language gap, which is a rather critical gap of the digital divide for Africa.
As earlier indicated, I had a major career breakthrough as a consulting engineer in 1995. I got the commission to design and build the biggest private television station in West Africa; the Africa Independent Television (AIT). I completed the project in 1997 and concluded that I had made my point in engineering and I should now concentrate on my passion to bridge the language gap of the digital divide for Africa. I therefore decided from then on to concentrate on developing the resources needed for the development of Human Language Technology (HLT) for African languages. I perceived human resources as the most important of these resources and so I decided to build a structure that will support the development of the human, intellectual, hardware and software resources. That was the thinking behind Alt-i which was presented to the public in 2004 as part of the West African Languages Congress which held at the University of Ibadan, Nigeria.
The first thing I did was to formalise my own knowledge of HLT by registering for a PhD in the University of Ibadan. This enabled me to support the University of Ibadan and the University of Lagos by teaching Artificial Intelligence and thereby introducing elements of Human Language Technology to the programs of the two institutions. Even though neither of the two universities has a formal programme in Computational Linguistics, Human Language Technology or Natural Language Processing till date, through the support of Alt-i, between both universities, five PhDs in these fields have been produced within one decade.
I count myself lucky to have been there at launch in 2004, and I can say that it played a role in my — even if subconscious — attention to language technology as a possible destination of my linguistics skills/obsessions. But since I left Ìbàdàn in 2009, I didn’t hear much about ALT-i, which could be because I wasn’t paying attention. What do you remember as your biggest accomplishment since launch till date?
The objective of ALT-i was to quietly do the needful. Many people could not understand why I should abandon my successful broadcast engineering career to dedicate the rest of my life to African languages which as far as they were concerned had nothing to add to our future. The linguistics students in the universities I worked with saw me as an enigma and I think you too must have seen me as so in your student days in UI. However, I was convinced that my hunches were right. Towards the end of 2004, Alt-i got the IICD award of the United Nations Economic Commission for Africa (UNECA) for creating a Yoruba keyboard that could handle all the diacritics and efficiently for that matter. The award gave Alt-i important international visibility and attracted funding for our work.
Many other awards came later, including the Silent Achievers Award of the Change A Life Foundation. It was presented by the then Governor of Lagos State HE Babátúndé Raji Fásọlá and I seized the opportunity to tell him that our work had been so far funded by foreign grant-making organisations. He responded by making Lagos State the first and still the only local entity that has funded our work. The biggest achievement however, is that we have provided support for the University of Ibadan and the University of Lagos to graduate 5 PhD’s in fields in which neither of them has a formal programme. In addition we have supported many other young Nigerians to gain admission for advanced degrees on Human Language Technology, Computational Linguistics and Natural Language Processing in universities abroad with scholarships.
We are therefore building a manpower pool for the development of a formal programme in any of these courses in a Nigerian university in the near future. These young people have undertaken many important projects like Yorùbá Speech Synthesis, Yorùbá Speech Recognition, Hausa Spell Checker, Automatic Induction of Igbo Morphology and many others. My first Ph,D student modelled the English writing styles of 5 West African countries. Her computer model could tell the country of origin of the writer of any document. This was in response to a need to provide some scientific evidence to the campaign of the then Nigerian Minister of Information (Mr.Frank Nweke Jr.) against the global notion that Nigeria was the sole source of cybercrime mails. Even though most of the projects address foundational technologies, we also have many products to show for them. We are proud to have motivated and facilitated these young people. These have always been and remain our objective and we continue to achieve them. I should also add that we provided local support to Microsoft Corporation for the localisation of Windows and Office Suite for Nigarian languages Hausa, Igbo and Yorùbá. We worked on Windows Vista, Windows 7, Windows 8 and Windows 10. The Nigerian languages with the three highest number of speakers (Hausa, Igbo and Yoruba) can now be typed on Windows in standard orthography..
If you have not heard much since you left Ibadan, it is because we have continued to pursue our objective of quietly doing the needful. It was so quietly done that you might not have realised that the person who did the technical part of the Yoruba speech synthesiser on YorubaName.com is from the ALT-i stable. She has now obtained a Master’s degree in the field and works in AI for a major bank in the UK. As you might know, there are very many of these young and not too young ladies and gentlemen working in cutting-edge employment all over the world. So far, we have managed to retain the 5 locally produced Ph.D in various universities in the country and very soon, we should have enough human resources to build a programme in Human Language Technology, Computational Linguistics or Natural Language Processing in a Nigerian university. Slowly but surely!
As you look at the environment today, what do you think are the biggest issues in African language technology?
In my experience, the biggest issue is the status given to African languages by Africans. Till now, very few African academics see any sense in applying their intellectual resources to the development of technologies for African languages. Many continue to aspire for improved mastery of their colonial legacy languages at the expense of African languages.
Another major problem is the departmental territorialism in many of our universities. Some of the rather jejune arguments that have slowed down the development of formal programmes in HLT is that a language degree cannot be obtained from a faculty of technology and a technology degree cannot be obtained from a Faculty of Arts. And there goes the prospect of a degree programme in HLT. I have also encountered the same level of uninspired and distracting arguments about Computational Linguistics.
In a piece I wrote a few years ago, I observed that Siri (and other voice technology solutions) exist in languages as small as Swedish, Norwegian, and Danish, three languages that have joint populations just half of Yorùbá’s, yet no Nigerian languages exist on these platforms. Why do you think this is the case? It’s certainly not because of population or literacy in the language.
There is a person-dollar value of every language. The person-dollar value of a language is the product of the population of speakers of the language and the amount of dollars in the pockets of these speakers. Language is the channel through which all negotiations are made. There is a justifiable perception of a greater need to negotiate with someone with lots of money than one with little or no money. Unfortunately, however, this composite person-dollar value of a language depends to a large extent on the perceived importance of the language, primarily within the speaking community. If a people feel that they can engage in negotiations and transactions in a foreign language, no matter the size of their population and the amounts of dollars in their pockets, the man-dollar value of their language will tend towards zero as the perceived value of their language, primarily within the speaking community tends to zero. All the man-dollar value their language could have accrued goes to the language they prefer to speak. That is the case of Igbo, Yorùbá and many other Nigeiran languages apart from Hausa. Recall that Swedish is the language that produced Ericson phones with just 10 million speakers and Finnish is the language that produced Nokia phones with just 5 million speakers! The population factor can be easily overrated.
But then, over the last couple of years, a number of localization efforts have emerged to throw some optimism in. I speak, of course, about the Nigerian English accent on Google Maps & Assistant (2019) and the new Nigerian English words on Oxford English Dictionary (2020). Although I played a part in both, and I believe that they are very important, I have also tried to temper people’s optimism about what value they add, especially to the African language space. The Nigerian English additions to the OED, for instance, do nothing to help codify Nigerian English in our educational syllabus. What do you think we can do, at home, as linguists and people involved in educational policy, to move the ball forward? How do we convince those in the position of authority?
Well, you speak of your own experiences. Recall that I coordinated the localisation of Microsoft Windows and Office Suite for Hausa, Igbo and Yorùbá for Windows Vista, Windows 7, 8, 10 and their corresponding Office Suites. The uptake was very low and I do not think Microsoft was impressed. Our illusion about what we can achieve with English is both deep and baseless. As long as we do not demonstrate that our languages are important to us we shall continue to be feeble copies of the owners of the languages we use for our development communications.
Do you remember that until very recently, banks in Nigeria did not give the option of any Nigerian language on their ATMs? One even offered French without offering the more relevant languages! As long as we continue to think this way we are not likely to achieve any of our development goals because we are not able to communicate with and carry a critical mass along at any time.
As for how to move the ball forward and convince those in authority, my policy is to play my role to the best of my ability. I can only enjoin others to do so too.
Something I find fascinating now in hindsight is how little information there was about linguistics by the time I was applying for the course in 1998, and how little people still know about it today. When I told people that I worked as a linguist at Google, one of the first questions I was asked was whether I work on Google Translate. The second is usually how many languages I speak. Why do you think this is the case, and how can we improve the knowledge and study of linguistics? Or, better still, how can we change the perception of the opportunities that exist in the field?
I alluded earlier on to how my own crass ignorance of the value of the subject of linguistics hampered my work in language technology. A Nigerian professor of linguistics once said she was asked by a second-year student of linguists when they were going to start learning the languages that will turn them into linguists. This level of ignorance is not limited to linguistics. It is a manifestation of the even greater ignorance of the value of education. In Nigeria, linguistics is seen as one of the “lower-ranked” subjects that an art student deigns to study having lost the competition to study law and other so-regarded higher-ranked courses. The tragedy for such a student is that linguistics is actually science!
To change this perception, the linguistics curricula of our universities need to be modified and the admission criteria need to be adjusted. This will be the first step to making graduates of linguistics sufficiently equipped to do the work demanded of modern linguists. If students who fail mathematics and physics can hide in departments of linguistics, such important subjects as phonetics, phonology, morphology, syntax, and semantics may suffer and these subjects may be substituted with language and literature courses and the perception will subsist.
One of the things you mentioned during our panel at Aké Festival in 2018 was how your early education in Yorùbá helped your knowledge of engineering. Can you speak more to this, against the background of the prevalent perception that only English language education can create geniuses in diverse fields? The Ifẹ̀ Six-Year Primary Project seems to have become something we merely reference as a past success rather than a blueprint for the future.
I have the rare privilege of being born and raised by two parents who did not suffer any inferiority complex about being Africans. For that reason, they did not think they were giving us a head start by speaking English to my siblings and me at home. Even though they were both competent in English, they spoke Yorùbá to us at home and so we were brought up as proud Yorùbá children. For that reason, my siblings and I got a very good understanding of our environment and we were able to describe it with competence because we got the benefit of a language that was developed primarily to describe it. Hence, as an engineering undergraduate in my late teens and early twenties, it was easy for me to visualise complex phenomena by translating whatever I was taught in English to Yorùbá in order to be able to contextualise them in ways that made them real. I remember clearly my problem with the notion of entropy. I was just unable to wrap my head around it until I saw it as ìdàrú; which made it clear to me that it is a measure of disorderliness. Take for example the English word “poison”. To a Yoruba pupil, it makes no meaning whatsoever beyond its being a label to a concept. On the other hand, consider its Yoruba equivalent “májèlé” which every educated Yoruba can easily analyse as “that which you eat and you are not able to eat ever again”. So, when a Yoruba child confronts “carbon dioxide”, and does not know that the “di” means two, he/she is left to memorise the label and is thus taught to learn by rote.
That all we do to the Ifẹ̀ Six-year Primary Project is refer to it in academic papers extolling it as a success and still do nothing to appropriate its success is the greatest evidence that we do not understand the value of education to society. We treat academic outputs as a means for individual academics to move along a career path rather than a means to reevaluate and re-engineer society. Great pity!
Considering that the current educational policies under which we currently raise kids in Nigeria are deficient with regards to mother tongue education, and that because the foundation was set many decades ago and will take a lot (in terms of time and resources) to unwind, how do you think we can begin to fix that particular problem without causing a major disruption?
We lived for hundreds of years before we were exposed to Western-style education. When it was time for us to make the decision on the language in which to learn science and technology, we thought the scientists and technologists who would teach the relevant subjects were the experts who should decide the language in which we should learn. Now we know better. It is linguists and educators that have the required expertise. Now that we know we erred, should we continue with the error and cause all six-year-old Nigerians to first learn English before they can be taught mathematics which was not invented in English or the history of their forebears which was not enacted in English? If we take the right decision today, it will take a few years to translate the relevant texts, train the teachers and develop learning aids. If however we continue to play ostrich, we shall continue to make our children learn inefficiently through cognitive proxies rather than learn by relating directly with reality. We, the adults, will continue to make “English’’ statements such as “it doesn’t worth it” and “I appreciate”, and get excited each time the Oxford English Dictionary approves words like “severally” as meaning repeatedly and “sef” as an emphasis tag to a statement. Definitely, language is dynamic, but we must admit that the contributions we purport to make to the English language come from our struggles to domesticate the language instead of taking due advantage of ours to teach our children efficiently.
Indeed. I also wanted to ask you about Ifá, something friends and people genuinely interested in it as a global cultural and spiritual resource have asked me. Why do we not yet have it online in one giant repository, accessible to all? I’m half playing devil’s advocate here because I am a little biased towards the original oral medium. But the web has given us tools to preserve both oral/aural and textual elements of this body of work, so why do we still need babaláwos to tell us what Chapter 7 of Èjì Ogbè says?
I agree with you that the knowledge of Ifá needs to be more widely disseminated as a global human heritage. It contains deep science that has been unfortunately mystified. The multimedia facilities offered by modern ICT definitely offer synergy for both the literary and the oratoral values it offers..
However, I do not think you can cure yourself of malaria by reading Gray’s Anatomy. The Babaláwo are intellectuals in their own rights and they cannot be wished away from the knowledge produced by their collective faculty of scholars over centuries.
I want to go back one more time to television and the role they once played in documentation and language revitalization. As a child, I watched Ifá Olókun Asọ̀rọ̀dayọ̀ which introduced us both to Ifá and the power of storytelling and fable. Later, we saw Bassey & Co and later Cockcrow at Dawn, Icheokwu, and so many more. I’m afraid that the answer I’ll get when I ask what has happened to the tapes of these and many more shows on NTA is that they have been lost, dubbed over, or badly preserved. Perhaps there’s nothing to do to fix the past, but how can the future be different?
Well, you talk about Bassey and Company, Cock Crow at Dawn, Icheokwu and probably The Village Headmaster. What about Safe Journey; The Adventures of Alao the Master Driver (Alao and Shakey Shakey), what about Sir Macauley John and many others of the late 50’s and early 60s? We have lost a lot of the media materials we grew on and for that reason, we have deprived our children of the sources of some of the values we hold, which they cannot understand. The managers of our media facilities have behaved irresponsibly over the years. The effects of their philistinism on society are palpable even though it is not easily realised that this irresponsibility contributes in no small measure to some of our contemporary malaise.
Fortunately, modern ICT has democratised access to the means of media production and everybody can now do their own thing. We see this effect on the development of Nollywood. People like Frank Donga, and the person that produced the “NEPA is owing me money” video would not have had a chance on the restricted public media we ran, but public media now has to compete with them for attention. We should learn from the mistakes we made in the past and now use modern ICT to document our lives. Our children should not have to resort to archeology to find out how we lived in the information age. As I said earlier on, we live in times in which we can synergise both the literary and oratoral values of texts. Moreover, it is now possible to combine the permanence of writing with the immediacy of speech.
And finally, what do you see in the future for African linguistics, African languages in technology, African languages in general, and localization technologies?
Language is the last bridge of the digital divide. Having laid all the marine cables, deployed fibre, installed microwave links and taken advantage of legacy copper networks, there still remains the last six inches of the digital divide, that gap between the terminal equipment and the eyes of the user, which can only be bridged by use of the prefered language of the computer user. African linguists are needed to develop the necessary theories and frameworks required for developing the language technologies that can bridge these gaps in as many as possible of about 3000 African languages.
Language will be the last frontier in our march as humans to make machines understand us. African linguists should set these as their targets and leave the teaching of local languages to graduates of education and leave working in media houses to graduates of Mass Communication. They have more compelling roles in developing more convenient user interfaces for ICT. These require studious modification of curricular and admission criteria.
Thank you for your time, Uncle T.
It was nice talking to you.
_____
This is the fourth in this series of public conversations I’m having with friends, mentors, colleagues, scholars, and enthusiasts, about our common obsessions. It is preceded by Conversation with a Conlanger (Jan 11, 2019), Conversation with a Polyglot (Feb 1, 2020), and Conversation with a Philologist (February 19, 2020).
