Technical Report 330, c4e-Preprint Series, Cambridge
Zaha: a RAG-based question answering system for the urban environment in The World Avatar
Reference: Technical Report 330, c4e-Preprint Series, Cambridge, 2024
Associated Themes:
- A natural language question answering system for urban data
- Ontological knowledge representation of urban data
- Diverse output formats tailored for various urban applications
As urbanisation accelerates, the demand for data-driven approaches to urban planning and management becomes increasingly critical. Urban data, encompassing geospatial, environmental, and regulatory information, is pivotal in informing the decision-making processes. However, various challenges such as data silos and the technical expertise required to query these complex datasets hinder the accessibility and utilisation of urban data. The World Avatar dynamic knowledge graph addresses these issues by integrating urban data across diverse sources and formats. Nonetheless, the process of querying from knowledge graphs remains non-trivial for non-experts due to the need for proficiency in query languages like SPARQL. This paper introduces `Zaha', a retrieval augmented generation question answering system integrated with The World Avatar knowledge graph, enabling natural language queries for urban data. By leveraging knowledge graph technology, Zaha is capable of efficient data retrieval based on the relationships defined between entities. As such, Zaha enables users, including urban planners, to access and query complex urban data intuitively, bypassing the technical expertise barrier.
PDF (3.0 MB)