Using natural language processing (NLP) to enable peer support between smallholder farmers
Over 1 billion smallholder farmers produce 80 per cent of the world’s food, and four of the five most traded commodities on earth. Yet the vast majority lack internet connections and access to up-to-date information to help them solve problems or share ideas with their peers. The primary challenges faced by peer networks are co‑ordinating between users efficiently and matching needs to existing expertise within the network.
Wefarm is a free peer-to-peer service that enables small-scale farmers to share information via SMS, without the internet and without having to leave their farm.
Farmers in Kenya, Uganda and Tanzania use Wefarm to ask each other questions about anything related to agriculture, then receive crowdsourced bespoke content and ideas from other farmers around the world within minutes. Machine-learning algorithms then match each question to the best suited responder within the network, based on analysis of the content and intent. The project uses natural language processing (NLP) models that can identify three regional African languages – Kiswahili, Luganda and Runyankore – in addition to English. The questions can be asked in any language and messaging is free of charge. If farmers don’t have internet access, they can access Wefarm via SMS on their mobile phones.
The AI provides an efficient routing between farmers’ needs and rare human expertise that is uniquely suited to solving the problem. The NLP models used by Wefarm are some of the first to support regional African languages, which enables access and advice to the broadest possible group of users. As a result, Wefarm has grown to become the world’s largest farmer-to-farmer digital network, with almost 2 million farmers using it in Kenya and Uganda alone, sharing more than 40,000 questions and answers daily.
CONSUL platform & Alan Turing Institute collaboration