The MGSMT Inference Module, which extends the MGSMT parser, includes a procedure for inferring a Minimalist Grammar (MG) lexicon using the the Z3 SMT-solver. If you use MGSMT Inference Module, please cite the publication detailing the inference procedure:
@InProceedings{pmlr-v217-indurkhya23a,
title = {A Procedure for Inferring a Minimalist Lexicon from an SMT Model of a Language Acquisition Device },
author = {Indurkhya, Sagar},
booktitle = {Proceedings of 16th edition of the International Conference on Grammatical Inference},
pages = {35--58},
year = {2023},
editor = {Coste, François and Ouardi, Faissal and Rabusseau, Guillaume},
volume = {217},
series = {Proceedings of Machine Learning Research},
month = {10--13 Jul},
publisher = {PMLR},
url = {https://proceedings.mlr.press/v217/indurkhya23a.html}
}
To see the inference procedure in action, you can run the accompanying Jupyter Notebooks. E.g. given input consisting of Primary Linguistic Data (each entry being a sentence paired with a skeletal representation of meaning), by successively running the four Jupyter Notebooks, the inference procedure will be used to (incrementally) infer an MG lexicon, concluding in the following lexicon:
which in turn can, for example, yield an MG derivation for "John has seen someone who was eating icecream.", as shown below:
(See the associated publication for more details about this example.)
The MGSMT Inference Module requires and has been tested with the following software:
- Z3 v4.8.7 (compiled using Clang 4.0.1)
- Python v3.7.7 (compiled using Clang 4.0.1)
- PyGraphViz v1.5
- IPython v7.12.0
- pdfTeX v3.14159265-2.6-1.40.21 (TeX Live 2020)
- Compiled with libpng v1.6.37
- Compiled with zlib v1.2.11
- Compiled with xpdf v4.02