Language 'can' be a barrier, we want it to be a bridge connecting the digital divide, we believe having a programming language that can allow and help educators to introduce computer programming using Hausa can enable non-English speakers in the North to widen their horizon.
Wow…okay, where do I begin? I’m not a writer so please do forgive if I sound incoherent as I narrate my story. :)
In December 2013, as usual whenever I get the opportunity to go on the internet, I go straight to Google’s translation portal to check if they have added any Nigerian language because I’ve been developing a Natural Language Processing (NLP) program that translates natural speech in text format from English to Hausa and lo! And behold! I noticed that they have just added Hausa in that same month, more than 15 years after I began developing mine. I started a few years before 1999 when Machine Translation became Microsoft’s NLP group’s major focus and 2001 when Google started providing the translation services.
It is a good thing I was not getting my livelihood from my program, otherwise, it is easy for Google to build something forty percent of the way and release it before completing it (at the time their English to Hausa was not great), therefore sucking the air out of the room. They do not even need to “finish” it — the very fact they have made it and put it out there everywhere is enough to make a market dry up and my program’s users to flock to theirs. It will have enough functionality — and just enough — to get the job done and it will be free.
This got me thinking really hard about competition, and Google. Particularly — how do you compete against the biggest technological behemoth ever seen by man?
In my case, the project started almost two decades ago, fresh out of secondary school and with a whole lot of time on my hands. At the time there was virtually almost no easy internet access in this part of the world – so no access to information on what I was about to embark upon – I went ahead using Basic on my Atari 65XE and then switched to Microsoft Qbasic on a computer with a 150MB hard disk, MS-DOS running on a 486x based processor (a top notch configuration at the time).
Not long after, the program had enough code and logic that could translate sentences like he is a boy – shi yaro ne, she is eating – tana ci, his father bought him a car – babansa ya saya [ma] masa mota, and the big book is on the chair – babban littafin ne kan kujeran, it understood hundreds of words and tens of sentence constructs.
To have an idea of some of the tasks you have to program a computer to do when translating between two human languages, I will give you a very simple example, consider these two very short sentences in English and their corresponding translations in Hausa:
“she is a girl” – “ita yarinya ce” and “she is eating” – “tana ci”, in the first sentence “she” is translated as “ita” while in the second sentence “she” is translated as “tana”. Any person who understands both the English and Hausa languages will translate the two sentences in a split second, but most will take a few minutes longer to try to figure out and explain why the word “she” took two different forms. Have you figured it out yet? Well, that’s what you have to program into the computer.
To cut – a more than a decade — long story short, my program reached the memory limit of the Qbasic interpreter, coupled with other factors resulting from added responsibilities in life — as I was only doing this somewhat as a hobby – the project took a back seat, until years later when I had access to more powerful, versatile technologies like VB.NET, MS Access Database, computers with more memory and powerful processors and Google/Bing Search (for quick access to information about Natural Language Processing) and also coupled with the potential of a project like this – I decided to further develop the program.
Competing with Google on a technological level is incredibly hard — it’s not impossible, just hard. They have so many PHDs and engineers, they have thousands and thousands of servers, they are richer than many countries – e.g. if you take the 2010 Gross Domestic Product (GDP) figures from the International Monetary Fund’s (IMF) World Economic Outlook Database, you will find that the combined GDP of the 28 poorest nations in the world is less than the $29.3billion that Google made in 2010.
But rather than dampening my enthusiasm, it almost instantly injected a whole new burst of energy – why? Well, I reasoned if Google is interested in building machine language translation algorithm for Hausa of which I’m also doing and my results standing well against theirs in some cases at the moment, and then I must be on the right track all along building something that might just be useful. Considering that, to date, according to my research, Google is the only one that is offering translation to Hausa.
I have decided to expand the road map of the project to include translating to more Nigerian languages, it’s not far-fetched but it’s definitely a long way to go because a project like this needs the program to understand a lot of language data and I think it’s something worth trying.
Perhaps, someday I might get the chance to showcase my program at a Technology Fair or better yet get the chance to participate in and win one of those international machine language translation competitions or maybe even succeed in developing a full blown near-fluent English-Hausa machine translation algorithm to serve as a back-end for Siri, Cortana or even Google Now before Google finally succeeds in building a billion word Hausa text corpus thus sucking the air out of my program. And my guess is eventually Google will finally succeed in building a billion word Hausa text corpus and suck the air out of my program, but before then thanks to Google for budging into my niche, I get to compete – albeit on a personal level — against the best and enjoy doing it.