Bitville

Deep neural networks to assist human learning

AI and machine learning has made notable progress over the past few years. Recently, the utilization of deep neural networks, also known as deep learning, has gained particular publicity within the field. Numerous interesting and intriguing research papers have been published lately regarding how deep learning methods can support in human learning. Companies are interested in evaluating the value and quality of various learning programs, and observing the skill levels of their learners. In the Spring of 2017, Bitville began a research project that uses machine learning and AI to evaluate the course quality and student skills levels in an automated way. The research project is partially funded by Tekes, and belongs to the Team Finland Augmented Intelligence campaign. Tekes is a governmental expert organization for funding research, development and innovation in Finland.

The solutions will be based on deep neural networks, and focus on:

Evaluating the course quality through automated analysis of textual feedback received from the students
Automated assessment of the learner's skills by analyzing the student interactions with the learning content
Automated assessment of written essays by using memory augmented neural networks

The Solution

By combining the three applications above, we have the potential to provide unprecedented tools for evaluating the quality and value of learning programs and also learner skills. The solutions are based on using a wide array of deep learning technologies including recurrent neural networks and memory augmented neural networks. The development and simulations are performed by using the TensorFlow deep learning software library, which is currently the state-of-the-art tool for developing and analyzing the deep neural networks.