Truthfulness detection plays an important role across a wide range of domains and experts have framed heuristics to help themselves understand deception. In such a domain like medical diagnosis, where patients lie to doctors, doctors have learned through experience which are the relevant characteristics to pay attention to, when speaking to patients, in order to understand whether they are being objective and truthful or they are telling untruthful information. In this work, we develop a framework for the creation of a data set for truth-telling recognition and we create a statistical model that, over time, learns characteristics and features correlated to truth-telling for a specific subject.