The evolution of information technology caused an expansion in the amount of data available on the internet. Moreover, such developments demanded that new tools were created to allow processing at high velocity, trying various informational sources. In this context, in flocking to the three V (Velocity, Variety and Volume), emerged the phenomenon called Big Data. From the emergence of this phenomenon, the need to generate new architectures that allow that users, enjoy the high volume of data spread throughout the Web. One way to improve the processes carried out, insert the question of semantic information processing, in which the use of domain ontologies can expand as computational agents interpret the meaning of the data. Thus, this paper aims to present a proposal for architecture that places the elements of Big Data and semantic, seeking to insert a model that is adapted to the current computing needs. As proof of concept performed the implementation of the architecture, exploring the question of scientific research, where a user is assisted to find relevant information in academic databases. Through the implementation, it was found that the use ontologies in a Big Data architecture, significantly improves the recovery of information performed by computational agents.