This report presents a graph partitioning approach given a set of constraints to resolve coreferences. Coreference resolution is the task of determining which referring expressions in a discourse refer to the same entity. Coreference resolution is a natural language processing task which has a direct effect on the field of Text Mining and its related areas such as Information Extraction, Question Answering, Summarization, Machine Translation. This report summarizes the research done in coreference resolution and presents our machine learning graph-based system and our baseline with preliminary results in comparison with other machine learning systems in the state of the art.