Despite the ever-increasing amount of information humans need to handle in modern society, the learning methods with which we teach and absorb knowledge have not adapted much over the past decades. The field of intelligent tutoring systems (ITS) tries to overcome this issue, but is still in its infancy. The goal of the project is to develop, implement and validate a neural-assisted, machine-learning-based ITS, that uses information from physiological signals to adapt the learning schedule to a user's individual learning capacity. In particular, decoding item-based memorizability from multi-modal EEG, NIRS and eye-tracking data will be analyzed by employing tools from deep learning – a relatively new field in machine learning that has had tremendous success over the past years. We are planning to decode the learner’s attentional state in real-time in order to help the tutoring system decide the pace and difficulty of the learning schedule in order to improve second language learning efficiency.