This is the repository for the EMNLP 2023 paper:
Information Value: Measuring Utterance Predictability as Distance from Plausible Alternatives
Mario Giulianelli, Sarenne Wallbridge, and Raquel Fernández
We provide the code for each set of experiments in the paper (see code/notebooks), as well as for generating the alternative sets used throughout the paper (scripts/generate). The alternatives themselves consist of 1.3M generations; we publish them here.
We're working on making this repository more user-friendly. Should you have any questions in the meantime, please reach out to mgiulianelli@inf.ethz.ch and s1301730@ed.ac.uk.