Lesan

The Web is one of the most important inventions that gives billions of people the ability as users and contributors, to connect with other people, access information and improve their lives and communities. For millions of people in Africa, the lack of relevant content in their language is a major challenge to fully leverage the Web.

In Ethiopia, a country with over 110 million people, Wikipedia is one of the top Websites used as a source of information on the Web. Essential information such as articles on top deadly diseases in the country is simply unavailable in local language editions of Wikipedia. In fact, the coverage of Amharic, Oromo and Tigrinya Wikipedia combined is less than 1% of the English Wikipedia. These are languages with over 70 million speakers. One of the top search queries on Google from this country is “translate”.  Language barrier is a pain that prevents millions of people from leveraging the Web.

Lesan’s mission is to open up the Web’s content for millions of people in their language. We do this by building machine translation systems. Currently Lesan has built state-of-the-art machine translation systems to and from Amharic, Tigrinya and English. 

Our machine translation pipeline solves the key bottleneck by leveraging online and offline data sources, a custom optical character recognition system for Ethiopic and an automatic alignment module that are powered by deep-learning.

Lesan has served over 90,000 users with more than 12 million translations so far. Many people use Lesan to translate news, health related information and educational materials form English into Amharic and Tigrinay. There are users who translate content from Amharic and Tigrinya into English to post on Social media and other outlets for a wider reach. For instance people have used Lesan to translate local content into English about the ongoing war on Tigray, Ethiopia.

Asmelash Teka Hadgu the founder and CTO comes from Tigray. His PhD research is in the area of machine learning where he designed algorithms that leverage the social web for applications in scholarly communication, crisis communication, and natural language processing for low resource languages.

Abel Aregawi our project manager is from Addis. He holds BA in Computer Engineering and is an MBA graduate with 4.0 CGPA. Prior to Lesan, he managed several projects in the tech industry.

They both know the pain many students have in understanding content because of language barrier. There's a lack of books and other educational materials in local languages. Together they are teaming up with local contributors to empower them with our state-of-the-art machine translation system to perform massive translation of locally relevant content.

Meaningful connections with partners and promoting Lesan.

Our solution significantly improves the state-of-the-art results for machine translation in low resource languages. We achieve this by leveraging offline and online data as well as taking advantage of cutting edge research in machine learning in our pipeline.

Our goal is to serve 100 million people in their language over the next year. We will achieve this by building machine translation systems for more languages in Africa. This will have transformational impact by making content on the Web available for millions of people in their language. We will also enable companies to enter these markets.

Two important metrics for Lesan:

- Number of people and companies using our machine translation service

- Volume of translations

We use data from unconventional sources and the power of machine learning to create state-of-the-art machine translation systems that power our service.

Our system, Lesan – Machine Translation for Low Resource Languages has been peer-reviewed and published at NeurIPS 2021 as a demo. This is a premier conference for machine learning research.

We value diversity, equity and inclusion. We are actively seeking to recruit talented individuals especially from historically marginalized communities.