diff --git a/README.md b/README.md
index 81d9b072f80ceb05af5a23ae30d768b754de1160..2e2b83f18ecb587e863daad5230e30a71a033e58 100644
--- a/README.md
+++ b/README.md
@@ -2,6 +2,7 @@
 ##### The two BN models are initialised by using the data collected from the human-therapist patient interactions and the therapist expertise in the form of questionnaire. The models are then refined session after session when the robot interacts with the patient. Therefore more accurate models are estimated.
 
 
+
 #### Package:
 - **bn_model_template** folder contains the models skeleton of the two models in bn_learn format
 - **bn_functions.py** and bn_variables.py are auxiliary classes for the bn_learn lib
@@ -10,3 +11,4 @@
 It returns a sequence of episodes defines as <current_state, assistive_action, next_state> that will be used as input for learning the reward function specific for that patient and therefore the degrees of assistance they need.
 
 
+#### NB: in order to build the BNs that models the therapist and the patient we used the [bnlearn](https://pypi.org/project/bnlearn/) library