A Framework for Cleaning and Integrating Multimodal Data in Digital Cultural Experience Research

The rapid digital transformation of cultural institutions has generated increasing volumes of complex data related to visitor behaviour, emotional engagement, and interaction with cultural content. Museums and cultural organisations increasingly use digital technologies, sensors, and interactive platforms to collect data about visitor experiences. However, these datasets are often heterogeneous and affected by issues such as missing values, measurement errors, inconsistent formats, and fragmented data structures, which limit their analytical usability and reliability. This paper proposes a structured methodological framework for cleaning and integrating multimodal datasets generated in digital cultural experience research. The framework is discussed with relevance to project contexts such as META-MUSEUM and unLoc and focuses on transforming heterogeneous data into more reliable, validated, and interoperable forms that can support future analytical processes and artificial intelligence applications. The proposed approach includes procedures for data validation, error detection, artefact removal, treatment of missing data, and standardisation of variable formats. In addition, the paper outlines a data integration process based on metadata harmonisation, format alignment, and the creation of consistent data structures that enable the combination of multiple data sources. The study highlights the importance of robust data governance practices for improving dataset quality, reproducibility, and analytical value. Its contribution is primarily methodological, offering a literature-informed framework for approaching multimodal data preparation in a more structured and reproducible way.