Probabilistic Pragmatic Modelling of Elliptical Corrections Using First Order Logic Semantic Representations

For the project report see the file report.pdf.

Execution: cd code && python3 court.py.

The figure shows what a pragmatic/smart listener infers the belief of a speaker who just made some remark to be, by taking into account their own world knowledge and considering the speaker's possible goals.

---

Dependencies

• nltk

• nltk external prover command (e.g. ResolutionProverCommand)

• torch

• pyro

• matplotlib

---

search_inference.py, HashingMarginal and inspiration taken directly from official pyro github examples The below is taken from the official pyro github examples

Rational Speech Acts (RSA) examples

This folder contains examples of reasoning about reasoning with nested inference adapted from work by @ngoodman and collaborators.

• generics.py: Taken from Probabilistic Language Understanding
• hyperbole.py: Taken from Probabilistic Language Understanding
• schelling.py: Taken from ForestDB
• schelling_false.py: Taken from ForestDB
• search.py: Inference algorithms used in the example models. Adapted from Design and Implementation of Probabilistic Programming Languages
• semantic_parsing.py: Taken from Design and Implementation of Probabilistic Programming Languages