ai for smallholder farmers

Wefarm harnesses artificial intelligence (AI) to empower human intelligence (HI). This means, all machine learning at Wefarm aims to supplement, augment, and identify the knowledge, skills, and achievements of our farmers. As we work towards connecting every small-scale farmer on earth, we’ll only get better at matching farmers with issues to farmers with the best solutions.

Our primary challenge is effectively connecting farmers who might be hundreds or thousands of miles apart and who rarely have internet access. That’s why Wefarm works over SMS – allowing anyone with a simple mobile phone to benefit from our network. Thanks to our unique machine learning technology, farmers don’t need English language proficiency to use Wefarm. Farmers can ask questions in any supported language and Wefarm's machine learning algorithms then match each question to the best suited responder.

Wefarm is already helping users overcome barriers to communication. It understands variation in spelling and punctuation as well as nuances in dialect and literacy. It even understands typos and works out what users mean to say. Our technology now give us the flexibility to easily add further languages as we move into new markets. While most data scientists in the world are focusing on the English-speaking western world, the data scientists at Wefarm are committed to using our skills to provide non-English speaking communities in emerging markets with better solutions to their every day problems.

algorithms for us all

frequency (1).png

Our Natural Language Processing (NLP) libraries are the first in the world to have created models which identify three regional African languages – Kiswahili, Luganda, and Runyankore – in addition to English. This means, unlike with other platforms, Wefarm users don't need proficiency in English and this places Wefarm at the forefront of breaking down the language barriers to technological development and advancement.

We’ve built from a system that originally understood only simple commands to a fully conversational interface (as used in chatbots) capable of naturally understanding the intent and content of messages and responding appropriately.

To do this, in each SMS we identify a few key features:

  • Language: The languages we support vary by country and we are looking to grow the number of dialects all the time. The Wefarm platform is highly scalable to different languages which will help take our service to farmers around the world.

  • Topic: By identifying the topic we can not only better direct the message to the relevant expert farmers, but on an aggregate level, we can identify macro trends for example—the effects of climate change and disease on small-scale agriculture.

  • Intent. Is the user asking a question or sending a response to an existing question? Are they trying to join us? Or are they sending us their name?

  • Spam: We ensure we do not send inappropriate questions or responses to other users.

Currently we ask each question to an average of around 20 users and users can expect a response inside 16 minutes, on average.