A unified semantic sports concepts classification as a key device for multidimensional sports analysis

Sports analysis tends to be more advanced and sophisticated in the last years, focusing on very detailed aspects of the game in order to result groundbreaking insights. New terms and expressions are shown up, corresponding to novel concepts and elements of sports. The diversity of terms used from various competitions and the media leads to a complexity for automated methods that aim to exploit textual sports sources for their analysis. This work searches, extracts and collects all the distinct terms that are used by major basketball competitions websites, via NLP techniques to create a generic classification of concepts related to basketball actions. The generated competitions vocabularies, as well as the generic vocabulary are transformed into SKOS hierarchical classification schemes and then linked to each other through their similar terms that correspond to the same actions. In this way, the generic basketball actions classification includes all the knowledge from the leagues vocabularies, the concepts and the synonyms from each competition in a hierarchical structure, forming a key device of comparability and aggregation. The use of the unified sports concepts classification could benefit the automation of multidimensional sports analysis through various competitions, years and media resources.