Code for the EMNLP-IJCNLP submission: "Learning Latent Parameters without Human Response Patterns: Item Response Theory with Artificial Crowds"

This zip contains the following files:

- README.txt: this file
- appendices.pdf: Appendices to the submission.
- fit_irt_svi.py: Python script to fit IRT models from response pattern data. Input data for the IRT model should be a single CSV file of the format: training_set_size, noise, pairID, response. 
- models/irt.py: Python file that implements the vague prior and hierarchical prior Rasch IRT models. 

Code has been tested with Python 3.6. Models require Pyro (http://www.pyro.ai) and Pytorch (http://www.pytorch.org)
