This web directory contains the Python scripts needed to train the computational model and generate model predictions, as well as the ratings collected for model evaluation.
The scripts used commands from the freely available DISSECT toolkit. The full DISSECT code, along with a tutorial, can be found at http://clic.cimec.unitn.it/composes/toolkit/

File list:
- FRACSS_training.py: the code used to train the FRACSSs, that represent affixes in our compositional model
- FRACSS_predictions.py: the code used to generate model predictions, both in quantitative and qualitative terms
- novel_words_plausibility.txt: tab-delimited file including plausibility scores for 600 novel words
- novel_words_neighbors.txt: tab-delimited file including the ratings for the experiment on novel-word neighbors (as predicted by the model)
- derived_words_ST.txt: tab-delimited file including semantic transparency ratings for 900 words (also used in Lazaridou et al., 2013)