A hybrid approach for decoding, reading and reshaping marginalized public spaces via ANT and AI knowledge graphs Une approche hybride pour décoder, lire et remodeler les espaces publics marginalisés via des graphes de connaissances IA et la TAR

This is a dataset repository for manusript titled: "A hybrid approach for decoding, reading and reshaping marginalized public spaces via ANT and AI knowledge graphs" ABSTRACT Urbanization in many developing countries often leads to fragmented planning and unsustainable governance, exacerbating climate vulnerability and social exclusion. This study reframes public spaces in marginalized communities as dynamic sociotechnical systems of agencies shaped by complex assemblages of environmental, social, economic, institutional, cultural, and technological actors whose interactions generate place-based resilience or fragility. In light of the latest advancements in big data and data-driven design tools, this study seeks to develop an ontological digital-hybrid framework aimed at deciphering these complexities and transforming them into a sustainable revitalization model by ontologically integrating the principles and mechanisms characteristic of digital public spaces. Using the El Baten district in Bousaada city, Algeria, as a concrete case study, this research employs a multidisciplinary hybrid approach that combines actor–network theory (ANT) as a robust methodological and ontological framework developed in science and technology studies (STSs)with AI-driven knowledge-graph tools to map, analyze, and visualize heterogeneous agencies and perform their relationships. The methodology follows a mixed-method approach that proceeds in two sequential steps. First, SWOT-based ANT analysis combined with multicriteria data analysis is used to construct the sociotechnical network of the marginalized public space and identify leverage points and vulnerabilities. Second, the network is improved by explicitly integrating the agency of technological objects via generative pretrained transformers (GPTs) through RAG technique, enabling the translation of sociopolitical complexity into an ontologically sustainable revitalization model. The findings demonstrate that the hybrid approach deconstructs latent relations, surfaces seldom-recognized actors, and generates actionable, context-sensitive revitalization scenarios that prioritize sustainability, local agency, and adaptive governance. The proposed model offers a replicable, multidimensional pathway for architects, urban planners, and policymakers seeking to foster resilient, inclusive public spaces in similarly marginalized contexts. By bridging ANT’s ontological insights with AI-enabled analytics, the study contributes both a conceptual framework and practical tools for sustainable urban revitalization of marginalized public spaces. keywordsANT, AI knowledge graph, public spaces, marginalized contexts, RAG technique, GPT-augmented UrbanKGs.