Knowledge base construction for the semantic management of environment-enriched built heritage: The case of Algerian traditional houses architecture

Traditional domestic architecture in Algeria reflects ancestral know-how which could serve as a model for new architectural achievements concerned with respecting the environment. This multidisciplinary knowledge covering different contexts (architectural, physical and environmental), is available but scattered through various sources of different formats (texts, tables, illustrations, etc.). Capturing such knowledge through an efficient knowledge repository is required. Knowledge base (KB) repositories have shown many significant contributions for managing the complexity, the variety of contexts and the querying of Built Heritage (BH) data. Creating a KB is a complex process that requires identifying and extracting the relevant knowledge from raw sources of information, then structuring, cleaning and integrating the extracted information into the KB repository. The management of these issues and their inter-dependencies are not always detailed in related BH literature. We propose in this paper a complete process for creating a KB dedicated to the built heritage consisting of Algerian traditional houses. Two main contributions are emphasized in this study: (i) the construction process of the KB is detailed at each step using two main suites: the ontology suite for identifying and structuring the relevant concepts, and the data suite that allows the flow of knowledge from the sources to the KB. (ii) the semantic variety of contexts of the input sources, their linguistic complexity and their format heterogeneity is handled all along the process, and is reflected at two levels: at the schema level (ontology model), and at the data level using automatic information extraction techniques. The definition of the KB includes fine-grained provenance metadata. Different results are proposed to evaluate the content and the quality of the KB. & COPY; 2023 Consiglio Nazionale delle Ricerche (CNR). Published by Elsevier Masson SAS. All rights reserved.