Fernando de Assis Rodrigues, Pedro Henrique Santos Bisi, Ricardo César Gonçalves Sant’Ana Identifying semantic characteristics of user interaction datasets through application of a data analysis Abstract In evaluating a decision, any fact analyzed needs to receive inputs from multiple data sources - structuring, integrating, storing, and processing collected data into an output that supports a better understanding of the fact from data, allowing new dimensions of analysis. The goal of this study is to identify the semantic characteristics of data attributes at the moment of collection, from dataset structures found in the data export interfaces of user interaction analysis tools, in Internet communication channels, and in web analytics data tools involved in scientific journal management, through the application of a process of data analysis and data modeling techniques. The research was delimited to exportable datasets available in interfaces from Open Journal Systems, Google Analytics and Search Console, Twitter Analytics, and Facebook Insights. An exploratory analysis approach was adopted to identify characteristics regarding how data are made available and structured in these data resources. Entity-Relationship Modeling concepts were applied to design and store data collected from services, resources, datasets, and attributes. In addition, the collected data was processed into another data structure, adopting the online analytical processing cube as a three-dimensional representation of elements, to facilitate analysis from different perspectives. This data analysis identified semantic dissonances in definitions of entity attributes, which may interfere with the process of developing relationships between attributes from different datasets, reducing the potential of interoperability. Introduction The use of data is part of the decision-making process in several fields, such as in education (Ikemoto & Marsh, 2007), industry (Reddy, Srinivasu, Rao and Rikkula 2010), management (Goodwin and Wright 2014), and science (Turban, Aronson and Liang 2004), among others. In evaluating a decision, any fact analyzed needs to receive inputs from multiple data sources – structuring, integrating, storing, and processing the collected data into an output that supports a better understanding of the fact from data, allowing new dimensions or perspectives of analysis (Inmon 2005; Kimball and Ross 2011; Reddy et al. 2010; Turban et al. 2004). For example, an evaluation of interactions between users and scientific contents in a publisher's web domain may be analyzed by service holders from the outputs generated in a process of collecting data regarding users’ interactions with their communication channels, structured into a data warehouse: a “[…] subject-oriented, integrated, time- variant, non-normalized, non-volatile collections of data that support analytical decision-making” with “[...] access to all information relevant to the organization, which may come from many different sources, both internal and external” (Turban et al. 2004, p. 236).