The social networks are a rich source of data and have been used to promote or organize cybercrimes that affect the real world. Because of this, the law enforcement agency are interest in the crucial information that can be get on this sources. The amount of information and the informal language which is used to spread information makes the Natural Language Processing (NLP) and excellent tool to make analysis over post in social media. That is why, in this proposal an architecture with three NLP models are integrated to provide an exhaustive analysis from open sources like social media. This analysis extract entities from the text, identifies clusters of users and their respective polarity, finally all of the results are related in a graph database. This architecture was under test using data from a real scenario in order to determine their feasibility.