Africa-Press – Eritrea. To do and to think are two different things. Your willingness to do what you need to do, pay the price needed, go the distance required, and make every possible sacrifice demanded is the measurement of how badly you really want to achieve your goal. If you believe you are destined to be successful, and you hold on to this belief no matter what happens, then there is nothing that can stop you from becoming that great success. Eng. Abiel, Eng. Fanuel, Eng. Meron and Eng. Siem are our invited guests for today’s Q&A column. These engineers are typical examples of those who are willing and working to become tomorrow’s great success. These engineers were able create the Tigrigna Text to Speech Engine (TTSE). The interview follows:
My name is Meron Mehreteab. I studied my secondary education in Asmara and headed to Sawa with the 29th round. I scored 3.4 on my matriculation exam and joined the Computer Engineering department in Mai-nefhi College of Engineering and Technology. I completed my college education with distinction.
My name is Abiel Amanuel. I studied at Barka Secondary School in Asmara and went to Sawa with the 29th round to complete my high school education and scored 3.6 in the matriculation exam. The same as my friends, I joined the department of Computer Engineering and received my first Degree from Mai-nefhi College of Engineering and Technology.
My name is Fanuel Teklemariam. I studied at Asmara-Hafeshawi Secondary School and headed to Sawa with the 29th round. I got 3.4 in my matriculation exam and joined the department of computer engineering and finished my five year studies there.
My name is Siem Haile. I studied at St. Mary Secondary School and went to Sawa with the 29th round to complete my national service and secondary education. I scored 3.8 in my matriculation exam and joined the College of Engineering and Technology.
Definitely! With the growing technological advancement, I always imagined being flexible in the work I perform. Therefore, I considered studying Computer Science as a way to achieve my goals, and I believed it could rhyme with what I always wanted.
There was an ICT teacher in my neighborhood who used to show me some computer applications back in my high school days. That grew up with me, and I always thought of joining the department of computer engineering. Hence from all the options at the College of Engineering and Technology, I had to choose only one and that was computer engineering. Nowadays, computer engineering is becoming influential as technology is trending and that’s why I chose it as it is more applicable.
When I was in high school, I started to realize that almost everything was being computerized and came to be fond of it when I went to Sawa. A demonstration was given by my teachers which left an impression on me. Not only that, but my parents were also very supportive when I told them that I was going to study computer engineering. With their support and inner ambition, I finally joined the department.
My uncle, my cousin and my two siblings are computer engineers. Having such influences around me was a simple reason for my inclination towards computer engineering. Especially my sister played the biggest role for me to join that department.
In fourth year, we took an introductory course on Artificial Intelligence (AI) and we were more motivated when we started fifth year. At the beginning, our plan was to create something related to robotics in connection to AI, particularly targeting the disabled, mainly those who can’t move their body or body parts. However, we didn’t go through with it because the finance and material required for the project was beyond our capacity. Therefore, we started brainstorming ideas on what we could do within our boundaries and cost-effective if possible. We then chose text to speech as a possible application to develop because it could somehow meet our criteria. To tell you something about the project, using AI, introducing Tigrigna to the rest of the world and allowing Eritreans to benefit from it was our aim. The project is mainly an AI which can imitate Tigrigna speech. We collected enough data and the sentences were prepared with their corresponding speech. Simply, our project was implemented in a way that raw text and audio data is provided to the end to end machine learning algorithm Tacotron proposed by Google (https://arxiv.org/pdf/1703.10135. pdf) in order to extract complex phoneme relationships among the different Tigrigna characters and then these extracted patterns can be used to synthesize speech of any given words.
Abiel: It is made to assist all members of the society equally. However, to arrange them in accord to the importance of the project to the users, those who cannot see or with sight problems are our priority targets. Those who cannot see, cannot read and to fill in what they are missing; this project can be helpful. Our second targets are Eritreans living abroad. These Eritreans, especially those having problems mastering the Tigrinya language, can be beneficiaries too. Our third target was the airport. This machine can assist travelers by notifying them their arrival and departure time in Tigrigna. All in all, this project is intended to help everyone in any way posible.
Our primary challenge was the fact that we were unfamiliar with such kind of technology because we only took one introductory course on AI and another on data mining (which we took when we were already halfway with our project) which was not enough to develop something out of it. Financial constraint was also another bottleneck for our project. As students we could only meet the expenses for the project to a limited extent. We knew from the beginning that we would be short of cash but it would be unfair to stop such a project because of money. The third challenge is the language. Tigrigna is a less sourced language and you cannot find the data for it easily. Another thing is we couldn’t record and we were left with no choice but to use audio books. However, we couldn’t find the editable written form of these audio books and we had to script it. Speech was crucial and we depended on it. Therefore, while training the artificial intelligence machine learning, high computational power was needed and this was challenging. In general, text to speech is more cost-effective to do as an enterprise, but we managed to tackle those impediments.
This project is quite big to be held only by students like us. Within that limited period and the data we were provided with (using 11,000 Tigrigna words) we were able to reach this stage. It is not deployable yet but at least we tried to demonstrate that it can be something big with time and financial support. Therefore, in the near future, we will try to make it deployable so that everyone could get the best out of it. To do so, we will need to collect more sentences and we are in the process of publishing it.
Our parents who gave us the motivation and financial support were the backbones for the success of our project, so we thank them. We also want to extend our gratitude to Dr. Yemane for his expertise advice. Writer Eden, studio-man Siem, Isaiah who was the narrator, our friends especially Henok Araia, Nahom Hagos and Belsabel Teklemariam who provided us laptops for setting our project’s python environment should be thanked too for their help. Last but not least, we would like to thank Mai Nefhi College of Engineering and Technology for the Internet access they gave us.