pynini 2.1.4

Finite-state grammar compilation

Pynini is a Python extension module which allows the user to compile, optimize, and apply grammar rules. Rules can be compiled into weighted finite state transducers, pushdown transducers, or multi-pushdown transducers. For general information and a detailed tutorial, see pynini.opengrm.org.

Pynini is primarily developed by Kyle Gorman with the help of contributors. If you use Pynini in your research, we would appreciate if you cite the following paper:

K. Gorman. 2016. Pynini: A Python library for weighted finite-state grammar compilation. In Proc. ACL Workshop on Statistical NLP and Weighted Automata, 75-80.

(Note that some of the code samples in the paper are now out of date and not expected to work.)

Installation instructions

Users can either install using conda-forge, or can compile the extensions and their dependencies from scratch.

Conda installation

Linux (x86) and Mac OS X users who already have conda can install Pynini and all dependencies with the following command

conda install -c conda-forge pynini

Source installation

Pynini depends on:

• A standards-compliant C++17 compiler (GCC >= 7 or Clang >= 700)
• The compatible recent version of OpenFst (see NEWS for this) built with the far, pdt, mpdt, and script extensions (i.e., built with ./configure --enable-grm) and headers
• Python 3.6+ and headers

Once these are installed, issue the following command:

python setup.py install

To confirm successful installation, run python test/pynini_test.py; if all tests pass, the final line will read OK.

Pynini source installation for the current version has been tested on Debian Linux 5.7.17-1 on x86_64, GCC 10.2.0, and Python 3.8.5.

Python version support

Pynini 2.0.0 and onward support Python 3. The Pynini 2.1 versions (onward) drop Python 2 support.

License

Pynini is released under the Apache license. See LICENSE for more information.

Interested in contributing?

See CONTRIBUTING for more information.

Mandatory disclaimer

This is not an official Google product.