Leveraging user-generated content to enhance heritage monitoring protocols and resources allocation through aspect-based sentiment analysis

Effective resource allocation for architectural heritage conservation requires reliable and up-to-date data to support proper decision-making. This paper proposes a novel methodology for the integration of reviews continuously generated by users with expert evaluations aimed at supporting decision-makers, which are based on periodic but infrequent site inspections and often fail to capture ongoing issues. This study utilises Natural Language Processing (NLP) and Aspect-Based Sentiment Analysis (ABSA) to improve the monitoring and management of architectural heritage. By leveraging user-generated content from Google Maps, this approach allows the integration of dynamic, real-time feedback into conservation strategies, addressing current limitations. The methodology was applied to data concerning a sample of 70 heritage buildings in Italy’s Marche region, whose state of conservation has been catalogued by the Central Institute for Cataloguing and Documentation (ICCD). The analysis highlighted significant discrepancies between expert evaluation and user feedback. While official assessments often miss recent changes or emerging issues, the ABSA-based tool succeeded in capturing insights such as recent restoration interventions or unnoticed maintenance needs. This approach introduces a methodology for a more inclusive, data-driven decision-making process, allowing resource allocation to be more responsive to ongoing conditions. By integrating expert evaluations and public inputs, the proposed methodology allows the adoption of more comprehensive perspectives, detecting blind spots in traditional monitoring systems and enabling more efficient management of resources.