On generating linguistic descriptions of time series

In this paper we provide a general approach for the concepts and processes related to the generation of linguistic descriptions of time series. As we will see, this approach consists of two main tasks, namely, a knowledge extraction task, which can be seen as a Knowledge Discovery in Databases (KDD) procedure, and a linguistic expression process. The presented approach incorporates as a core element a description model, which is based on three pillars: a knowledge representation formalism, an expression language, and a quality framework. In the paper, we also analyze the main tools and techniques that can be used regarding the mentioned tasks and pillars of the generation of linguistic descriptions of time series. Additionally, we provide a deep review of the main contributions in the area, which come mainly from the fields of Natural Language Generation (NLG) and Fuzzy Sets and Systems. The existing and potential contributions of fuzzy sets and extensions are discussed in detail. Together with the application of KDD techniques, we encourage the cooperation of the Fuzzy Sets and the NLG communities in order to provide a significant step forward in the development of systems for providing linguistic descriptions of time series data.