Classification methods for email: an interdisciplinary approach between Knowledge Engineering, Archival Science, and Data Science

Resumen

In the public or private sectors, email has been adopted as a tool for transmitting attached files or messages for internal and external communication. With more than a decade being used in corporate environments, its users started to adopt it as a tool to explicit knowledge of their activities and transactions. The general objective of the research is to classify an amount of email according to a predefined classification scheme established for Brazilian Federal Executive branch. It is an applied research, with qualitative approach and exploratory objective. It adopted the Proof of Concept (POC) method, performing the classification task using traditional text classification algorithms, as well as the vector representation by Bag of Words (BOW). A theoretical review on classification of texts used in this research considered the Knowledge Engineering, Data Sciences, and Archival Science areas. It was possible to infer that the classification has the same purpose in these areas of knowledge: to group an object to facilitate the retrieval of information and knowledge by its user, confirming that there is interdisciplinarity between these areas. The result achieved was an accuracy rate of 52,71% with kNN, SVM, MultiClassClassifier, and test model k-fold cross validation. It is concluded that the model induced in the research has the potential to be applied, in an experimental way, in email storage considered as a step prior to reliable digital repositories adopted as a resource for storage, management and preservation of an email message recognized as explicit knowledge.