Uliza

Offering readers access to internet services by voice and text in their native languages.

People around the world are getting internet devices, but their experience depends on the languages they know. With almost 90% of content in Western European languages, Uliza is a solution to the language barrier in internet services. We care about letting people use their voice in their mother tongue to participate in global information services.

Uliza brings the internet into African and Asian languages, in voice and text. We equip businesses and organisations with near-real-time voice recognition, translation, and chatbot tools, so they can offer their services in more languages, to billions more people. We do this by crowdsourcing transcribing and translating voice data, and matching questions to answers. A large team of remotely based freelance experts compete to complete tasks, giving us a turn-around time as low as five minutes. Businesses access our services online through our API and pay us per task completed. As we do this, we are building language datasets for training machine learning models to automate our services.

Internet services will play a large role in the future of work. The impact of access to information has been proven by economists to be significant and positive, particularly for rural development. We will give voice access to the internet to the students, farmers, and entrepreneurs of the future. Our positive impact gives us support from communities, donors, and impact investors, and this adds significant strength of our business model. Moreover, our crowd sourcing model is providing hundreds of part-time jobs for people all over Asia and Africa.

First, our freelance workforce is structured to deliver near real-time service for clients that need bots and voice recognition not yet served by automation.
Second, businesses can use our services in channels that interface directly with consumers, like calls, SMS, and WhatsApp.
Third, an API that allows clients to slot our service into already-built products making it easy to expand existing products into African and Asian markets with our API backing the language transition.
Lastly, we are focused on serving African and Asian languages, where we have a first mover's advantage.

We have built a platform and an API that allows crowd workers to logon to our platform and complete tasks for pay, such as transcription, translation, or matching an answer to a question.
There are automated quality checks and crowd-management tools built in to make the system efficiently scalable.

We also use perform early stages of natural language processing to prepare the data for machine learning, and identify frequently asked questions, which we will leverage to build automated voice recognition and translation in under-served African and Asian languages. These processes also allow us to onboard a new language quickly.

Our top three goals over the next year are:
Give 1 million people mother-tongue access to the services of our business clients.
Increase our language offering to cover: Kiswahili, Hindi, Urdu, isiZulu, isiXhosa, Marathi, Gujarati, and Hausa.
Secure investment to develop automated voice recognition software in those languages with our voice data.

Our vision is to include everyone in the benefits that come from being able to access the information available on the Internet in your mother-tongue. In 5-years time we will provide natural language processing tools, like automatic voice recognition, in African and Asian languages, so that organizations can include the 4 billion people who do not speak major Western European languages. This will revolutionize the way a large proportion of the world currently interact with the Internet and bridge the information asymmetry that exists today.

While Uliza has a broader impact of providing access to information to people who are not able to use the internet or do not speak a major Western European language, we measure our immediate impact on a project-by-project basis. With Precision Agriculture for Development and MercyCorp in East Africa, we are helping expand their agricultural services to reach farmers who do not speak English and who do not have the education to use their existing offering. We have also helped create hundreds of new part-time job opportunities for people across Africa and Asia, by offering payment to local language speakers.

By allowing people who have previously been unable to access the information available on the internet because of language and literacy barriers, Uliza is transforming how marginalized people can join the mainstream economy. Information is power for the people.

Furthermore, all our crowd experts are compensated through mobile money and online payment services such as PayPal, which has allowed us to hire workers all over the African and Asian continent to work for us and earn money while helping us build the technology that will bring automated voice recognition to their mother-tongue languages.

Janeth is one of Kenya's top young technical minds and is a technology analyst at Deutsche Bank and leads managing Uliza's technology needs.

Maity has a background in mathematics and finance and has worked as a data analyst and consultant in the United States and as an economics teacher in his native India. Maity manages Uliza's financial and legal affairs.

Grant is from South Africa, and his background is in development economics, curating data for professors at MIT, and applying technologies to developing country contexts.

1. Use crowd-management tool combined with automated natural language processing technology to provide businesses with transcription, translation, and chatbot services that are accurate, low-cost, and quick. We will outcompete traditional transcription services in Africa and Asia on cost and speed through our use of crowdsourcing and NLP technology. We will outcompete the products of the tech majors because of their consumer-focussed offering that is low-quality or absent in African and Asian languages.

2. We secure licenses to use the voice data that our crowd process to build automated language technologies using machine learning tools. We will secure sufficient quantities of data to develop business-grade voice recognition and translation technologies, in-house and in partnership. We will leverage both our language data and our automated technologies, by licensing, selling, or openly distributing them.

3. We will focus on business opportunities in strategic languages, such as Kiswahili, isiZulu, Hindi, and Urdu. Our backend tools makes it easy, and with low marginal cost, to replicate the process for additional languages, and we will use the revenues from the first languages we pursue to launch into the next.

We understand our marketplace in Africa and Asia well, and we have an excellent and encouraging pipeline that we want to serve in 2018. Our biggest bottleneck is technical human resources that is slowing us from deploying our technology quickly enough. We will use any future funding in bringing on a full stack developer and an additional web developer to customize our services for our clients. We would also like to allocate some funding in attending development conferences and hackathons to showcase our services.

Internally, we need more machine learning expertise.

Externally, our biggest risk comes from the tech majors (Google, Amazon, IBM, Baidu), developing high quality voice recognition in major African and Asian languages sooner than 3 years time, which would make our service redundant. We do not expect this to be the case for two reasons:
1. In African and Asia, they are not able to benefit from large volumes of publicly translated texts and transcribed audio, as they did for European languages.
2. We don't expect them to be as knowledgeable about how to collect their own data in our region.