Sphinx is a human-AI hybrid system that helps writers interactively create English reading passages. Our system utilizes natural language processing (NLP) models to help expert or novice writers interactively create English reading passages in an efficient, scalable manner.
These passages could be used in a variety of learning and assessment applications such as formative and summative assessments of reading and comprehension and real-time adaptive learning.
We utilize powerful NLP transformer models such as BERT (Liu 2019) to encode text data and automate writer tasks such as topic modeling, summarization, sentence recommendation, and paraphrasing.
The system is designed to make recommendations regarding passage content that are evaluated by human users before inclusion. This enables quality control as well as collecting training data for underlying machine learning models.
We believe such human-AI hybrid systems can be the best of both worlds by utilizing the reasoning ability of subject matter experts in delivering adaptive learning experiences as well as formative and summative assessments of English reading among other educational applications.
Gierl, M. J., Lai, H., & Turner, S. R. (2012). Using automatic item generation to create multiple‐choice test items. Medical education, 46(8), 757-765. https://doi.org/10.1111/j.1365-2923.2012.04289.x
Liu, Y. (2019). Fine-tune BERT for Extractive Summarization. arXiv preprint arXiv:1903.10318. https://arxiv.org/abs/1903.10318