- Harvard University and Microsoft.
- ALOSI (Adaptive Learning Open Source Initiative) is based on creating an open source adaptive engine powering individualized learning and assessment pathways.
- Collaborative-based Personalization
- Zachary A. Pardos, Assistant Professor UC Berkeley
- Data-Driven Syllabus Adaptation
- Data & AI as a translation between expert/novice learner models
- RNNs langue models adapted to course navigation sequences. Intended to pick up on patterns among successful learners.
- Course Recommendation Systems with Open edX Data
- Chinmay Nivargi, Stanford University
- Content-based Recommenders
- How similar is Course A to Course B?
Represent each course’s bag-of-words as a vector
TF-IDF (Term Frequency – Inverse Document Frequency) was used as the weighting for every word - Courses previously completed by user, Courses rated by user (if available), Any other course – user relationship data (e.g. topics, region, etc.)
AI Grading
https://edx-ora-2.readthedocs.io/en/latest/architecture/ai_grading.html
Training a classifier using a supervised machine learning algorithm.
Grading student essays using a trained classifier.
Both training and grading require more time than is acceptable within the request-response cycle of a web application. Therefore, both training and grading must occur asynchronously.
The architecture should not constrain the ML algorithm (or algorithms) used by a particular implementation. It should be possible to replace the ML algorithm with any supervised learning algorithm that produces a text classifier.
We also avoid constraining the particular task queue implementation. In principle, any task queue that provides basic reliability guarantees and a retry mechanism will work (see Entities).