Abstract:
The advent of Shipping 4.0 is driving a surge in the generation and accumulation of extensive maritime
shipping operation data. These data, integral to maritime digitalization, underpin critical
transformations and applications, including maritime big data analytics and AI-based solutions, for
smart shipping, autonomous shipping, green shipping, and the development of next-generation ports.
Despite these advancements, effectively harnessing the substantial volume of maritime data presents
challenges, primarily attributable to variations in data quality across diverse maritime systems. The
absence of thorough and consistent data quality examination and monitoring, especially the failure to
detect issues resulting in data mis-matching, along with the necessity for high data quality in the early
stages of AI application development, can significantly impact the efficient development and
confidence in artificial intelligence (AI) applications within maritime systems. This challenge is
exacerbated by the inherent complexity of data quality issues.
To address these challenges, we developed a maritime-specific data quality assessment framework
featuring reliability, flexibility, and traceability. This framework integrates a Syntactic-SemanticPragmatic (SSP) layered structure and multi-dimensional data feature representation. Within this
framework, data quality is stratified into three layers: the syntactic (conforming to metadata and
maritime data quality standards), semantic (degree of representing real-world entities considering
context information), and pragmatic layer (fulfilling specific application needs). In each layer, data
quality is assessed and described within five dimensions, namely precision, completeness, uniqueness,
and accuracy, with the consideration of examplar datasets in maritime systems. The inclusion of
maritime domain knowledge, such as maritime data standards, maritime databases, and maritime
applications, enhances the reliability to identify maritime-specific issues. The overall mutuallyexclusive and collectively-exhaustive framework provides adaptable integration capabilities, allowing
for customization to suit varying data and organizational needs. Moreover, the framework provides
quantitative evaluation of data quality, enables tracking throughout its lifecycle and promotes
transparency and accountability in data processing processes.
The effectiveness of the proposed data quality assessment framework is demonstrated through its
application to AIS data, which stands as a critical dataset within the maritime sector. Data entities from
different sources were analyzed, evaluated, and compared with this framework. By offering practical
insights into data quality realities and trends across different regions, the framework contributes
significantly to enhancing data quality awareness.
As one of the first maritime data quality assessment frameworks, our method seamlessly integrates
syntactic, semantic, and pragmatic considerations. This approach facilitates robust diagnosis of data
quality issues in the marirtime context and allows adaptable evaluation of data assets as well as
continuous monitoring of data quality quantitatively. Ultimately, our framework empowers the
development of reliable maritime AI applications, which is essential for the maritime digital
transformation process.