The analysis of scientific collaboration networks has contributed significantly to improve the understanding of the collaboration process between researchers. Additionally, it has helped to understand how scientific productions by researchers and research groups evolve. However, the identification of collaborations in large scientific databases is not a trivial task, given the high computational cost of the prevalent methods. This paper proposes a method for identifying collaborations in large scientific databases, namely, ISColl-Identification of Scientific Collaboration. Unlike methods that use techniques such as exhaustive comparisons of publication pairs, the proposed method produces satisfactory results with a low computational cost, thus providing an interesting alternative for the modelling and characterization of large scientific collaboration networks. To demonstrate the potential of the proposed technique, tests were conducted using scientific publications data registered in the Lattes Platform of CNPq, with the obtained results yielding excellent accuracy during the identification of scientific collaborations. Palavras-chave: Extraction and data integration. Information retrieval. Identification of collaboration. 1 Introdução The production and publication of scientific papers have increased considerably in recent years. The rapid proliferation of research publications on the Internet can be considered as the primary factor accelerating the distribution of this class of publications. Services such as digital libraries, social networks, websites and bibliographic repositories that act as a personal storehouse for an individual's scholarly or scientific productions are some examples of how the Internet has A method for the identification of collaboration in large scientific databases