pyfcg 0.2

A Python port of Fluid Construction Grammar

PyFCG

A Python interface to FCG and Babel, built on FCG Go.

Installation

The Python integration of FCG (aka PyFCG) is distributed as a pip-installable package. To install this package from PyPI, run the following command.

$ pip install pyfcg

Use

Once pip-installed, PyFCG can readily be imported as a module into Python programs. It is customary to define fcg as an alias for the PyFCG module, so that all functionality is available within the fcg namespace (functions and variables are thereby prefixed with fcg.)

You initialise PyFCG as follows:

>>> import pyfcg as fcg
>>> fcg.init()

Many FCG functions depend on the fcg-go precompiled binary, which you loaded (or download if necessary) by calling the init() function. On macOS, you will need to accept opening the application bundle by clicking "open" in the popup window.

Example: Creating an FCG Agent

The Agent class will usually be your main entry point. The idea is that an agent has a grammar (of type Grammar), which in turn holds constructions (of type Construction). In PyFCG, a grammar will always be tied to an agent. Upon creation, an agent is automatically initialised with an empty grammar, i.e. a grammar that holds zero constructions.

>>> demo_agent = fcg.Agent()

>>> demo_agent
<Agent: agent (id: agent-42) ~ 0 constructions>

>>> demo_agent.grammar.size()
0

Walkthrough Tutorials

A more in-depth showcase of PyFCG's functionalities can be found in the following interactive notebooks. Each of them walks you through the process of integrating PyFCG in a typical use case of FCG.

• Tutorial 1: Grammar Formalisation and Testing
• Tutorial 2: Learning Grammars from Corpora
• Tutorial 3: Modelling Emergent Communication

Documentation

The tecnical documentation of PyFCG is available at Read the Docs.

How to Cite

Van Eecke, P., & Beuls, K. (2025). PyFCG: Fluid Construction Grammar in Python. arXiv preprint.

@article{vaneecke2025pyfcg,
    author = {Paul {Van Eecke} and Katrien Beuls},
    title = {PyFCG: Fluid Construction Grammar in Python},
    year = {2025},
    journal = {arXiv preprint arXiv:},
    doi = {10.48550/arXiv.2505.12920},
}