Um estudo bibliográfico sobre ligação de entidades

Linking Entities (LE) is an important research topic that has recently attracted great attention from researchers. In the LE tasks, textual mentions found in natural language are linked to their corresponding entry in a knowledge database. This\udtask is challenging due to problems such as name variation, entity ambiguity, or because the entity mentioned may not exist in the knowledge database.\udGoals: To present the problems related to LE, its typical applications, as well as to synthesize their main approaches in the context of concept linkage.\udMethodology: Survey research with the current literature, for a detailed description of\udthe state of the art of the approaches in LE, as well as for the systematization and categorization of the identified approaches.\udResults: Most of the studies in LE divide this process into two stages: recognition and linking of entities. However, new proposals have unified these steps into a single process.\udConclusion: Although more complex, the new LE approaches allow us to capture the dependence between Entity Liaison and Entity Recognition decisions, minimizing errors and inconsistencies. Evaluations should occur in unified databases, considering the difficulty of comparing different database results due to the influence they have on testing outputs