Lacuna Fund - Open Data Sets & Use Cases

Under this initiative, a novel approach is employed by leveraging citizen scientists and IoT-based low-cost sensors to collect hyperlocal air quality data. This data is used to identify pollution sources and risk zones, facilitating targeted actions by regulatory authorities.To showcase data outreach, the project features the VAYU Android-based application and the VAYU citizen portal digital stack, which support targeted interventions and customized solutions backed by AI/ML algorithms. These tools potentially develop new approaches in air pollution management while reducing public investment costs.

The NaijaSenti dataset is the first large-scale human-annotated Twitter sentiment dataset for Hausa, Igbo, Nigerian-Pidgin, and Yorùbá, the four most widely spoken languages in Nigeria. It consists of around 30,000 annotated tweets per language (except for Nigerian-Pidgin), including a significant fraction of code-mixed tweets. These datasets are useful not only for sentiment analysis but also for hate speech detection. The open-source package consists of datasets, trained models, sentiment lexicons, and code. With more than 200 million people and 522 native languages, Nigeria is the most populous and linguistically diverse country in Africa, as well as the third most multilingual country in the world. The majority of the population speaks either Hausa, Igbo, Yorùbá, or Nigerian-Pidgin.

An AI-Driven Dataset for Nature-Positive Livelihoods and Forest Restoration in Eastern Himalayas.This open-access dataset and digital MRV (Monitoring, Reporting, and Verification) toolkit was developed to help communities, NGOs and policymakers in the Eastern Himalayas measure and manage the links between forests, livelihoods and climate resilience. The project was created in response to a persistent data gap in the Himalayan region where limited open data and difficult terrain have hindered effective monitoring of forest degradation and the impact of restoration programs. Using locally calibrated AI models trained on Indian biomass data, the system combines satellite observations with ground level socio-economic surveys to measure forest dependency, identify degradation hotspots and track restoration outcomes. All datasets ranging from community level livelihood surveys in Sikkim and Mizoram to forest biomass and land use layers in Assam are openly available. Together they form a replicable, open-source “digital public good” that can be used to train local AI models for forest monitoring, biodiversity accounting or impact assessment. By providing a scalable, context-aware dataset, this initiative enables indigenous start-ups, environmental NGOs and impact investors to quantify nature positive outcomes, align with IRIS+ indicators, and guide conservation investments. The goal is to make environmental monitoring inclusive and data driven, empowering local actors to track changes in forest carbon, identify erosion risks and measure the socio-economic benefits of restoration. Ultimately, this approach helps governments, researchers and entrepreneurs build locally adapted AI systems that support both ecosystem regeneration and sustainable livelihoods in fragile mountain regions.

This dataset provides continent-wide crop field labels for Africa, improving the availability and use of crop field boundary (parcel) maps. It contains 42,403 annotated geospatial polygons indicating the boundaries of individual crop fields spanning the years 2017-2023. This enables the development of automated field boundary detection systems for agricultural monitoring across Africa.