Library for manipulation of phonological distinctive features
Project description
DistFeat Python library
DistFeat is a Python library for manipulating segmental/distinctive phonological features.
Installation and usage
The library can be installed as any standard Python library with
pip
, and used as demonstrated in the following snippet:
In any standard Python environment, distfeat
can be installed with:
$ pip install distfeat
Note that the command above will install the pyclts
depency, but will not download
any version of the CLTS data by default.
Detailed instructions on how to use the library will be made available in the official documentation. Code documentation and test cases detail usage, along with the following section.
Showcase
Functionality is provided by means of a DistFeat
class, which will
automatically load the standard model upon instantiation:
>>> import distfeat
>>> df = distfeat.DistFeat()
The most common functionality, obtaining a dictionary of features for a
grapheme, is performed by the .grapheme2features()
method.
>>> df.grapheme2features('a')
{'anterior': True, 'approximant': True, 'back': False, 'click': False, 'consonantal': False, 'constricted': False, 'continuant': True, 'coronal': True, 'distributed': True, 'dorsal': True, 'high': False, 'labial': False, 'laryngeal': True, 'lateral': False, 'long': None, 'low': True, 'nasal': False, 'pharyngeal': None, 'place': True, 'preaspirated': None, 'preglottalized': None, 'prenasal': None, 'round': None, 'sibilant': False, 'sonorant': True, 'spread': False, 'strident': False, 'syllabic': True, 'tense': True, 'voice': True}
The .graphemes2features()
method will by default returning a dictionary with
boolean values, with sorted feature names. Arguments allow to skip the
truth value conversion, returning the strings used for their representation,
and to return a vector of values as a list.
>>> df.grapheme2features('a', t_values=False)
{'anterior': '+', 'approximant': '+', 'back': '-', 'click': '-', 'consonantal': '-', 'constricted': '-', 'continuant': '+', 'coronal': '+', 'distributed': '+', 'dorsal': '+', 'high': '-', 'labial': '-', 'laryngeal': '+', 'lateral': '-', 'long': '0', 'low': '+', 'nasal': '-', 'pharyngeal': '0', 'place': '+', 'preaspirated': '0', 'preglottalized': '0', 'prenasal': '0', 'round': '0', 'sibilant': '-', 'sonorant': '+', 'spread': '-', 'strident': '-', 'syllabic': '+', 'tense': '+', 'voice': '+'}
>>> df.grapheme2features('a', vector=True)
[True, True, False, False, False, False, True, True, True, True, False, False, True, False, None, True, False, None, True, None, None, None, None, False, True, False, False, True, True, True]
The operationally inverse method .features2graphemes()
returns a list of all
graphemes that satisfy a set of features and their values (which can be
provided both as truth values or as their strings). It is possible to drop
undefined values by means of the drop_na
argument.
>>> df.features2graphemes({"consonantal": "-", "anterior": "+", "high": "-"})
['a', 'aː', 'ã', 'ãː', 'ă', 'ḁ', 'a̯', 'e', 'eː', 'ẽ', 'ẽː', 'ĕ', 'e̤', 'e̥', 'e̯', 'æ', 'æː', 'æ̃', 'æ̃ː', 'ø', 'øː', 'ø̃', 'ø̃ː', 'œ', 'œː', 'œ̃', 'œ̃ː', 'ɶ', 'ɶː', 'ɶ̃', 'ɶ̃ː']
A minimal matrix of features needed to distinguish a set of graphemes can be
obtained with the .minimal_matrix()
method, which also allows to use
strings for truth values and to drip undefined values. Like in the
case of .grapheme2features()
, a vector
argument can be passed in order
to obtain a list of values. As expected, the
larger and more heterogeneous the set of graphemes, the larger the
number of features needed. The snippet below also used the auxiliary
tabulate_matrix()
function, a wrapper to the tabulate
library.
>>> distfeat.tabulate_matrix(df.minimal_matrix(["t", "d"]))
constricted laryngeal spread voice
-- ------------- ----------- -------- -------
d False True False True
t False
>>> distfeat.tabulate_matrix(df.minimal_matrix(["t", "d", "s"]))
constricted continuant laryngeal sibilant spread strident voice
-- ------------- ------------ ----------- ---------- -------- ---------- -------
d False False True False False False True
s True False True True
t False False False False
>>> df.minimal_matrix(["t", "d"], vector=True)
{'d': [False, True, False, True], 't': [None, False, None, None]}
The operationally inverse method to the one above is .class_features()
,
which provides a dictionary of features and values to constitute a class of
sounds from a set of graphemes. Note that, while possible, this method
does not drop undefined values by default. As expected, the larger and more
heterogeneous the set graphemes, the fewer the number of feature/value
pairs in common.
>>> df.class_features(["t", "d"])
{'anterior': True, 'approximant': False, 'click': False, 'consonantal': True, 'continuant': False, 'coronal': True, 'distributed': False, 'dorsal': False, 'labial': False, 'lateral': False, 'nasal': False, 'place': True, 'sibilant': False, 'sonorant': False, 'strident': False, 'syllabic': False, 'tense': False}
>>> df.class_features(["t", "d", "s"])
{'anterior': True, 'approximant': False, 'click': False, 'consonantal': True, 'coronal': True, 'distributed': False, 'dorsal': False, 'labial': False, 'lateral': False, 'nasal': False, 'place': True, 'sonorant': False, 'syllabic': False, 'tense': False}
A simple command-line tool for querying the database is also provided.
Experimental support for segment distance is available as well, as
demonstrated below. It requires the sklearn
library, which is
not listed as a requirement and, as such, is not installed by default.
As models and regressors are not cached, the training
phase might take longer than expected.
>>> df.distance("a", "e")
5.501464265353438
>>> df.distance("a", "u")
6.773080283814581
>>> df.distance("w", "u")
0.9799320477423237
>>> df.distance("s", "ʒ")
10.139607771554383
Changelog
Version 0.2:
- Added initial support for segment distance
Version 0.1.1:
- Added unround open-mid front vowels which were missing from the default model
- Added a model derived from Phoible
Version 0.1:
- First public release
TODO
- Allow to specify, check, and derive geometries
- Decide whether to have
.features2graphemes()
defaulting to boolean values (i.e.,t_values=True
) - Decide on how to specify undefined when using truth values, such as in
.features2graphemes()
(considering thatNone
cannot be passed as a value) - Extend the command-line tool to call most if not all functions
Community guidelines
While the author can be contacted directly for support, it is recommended that third parties use GitHub standard features, such as issues and pull requests, to contribute, report problems, or seek support.
Contributing guidelines, including a code of conduct, can be found in
the CONTRIBUTING.md
file.
Author and citation
The library is developed by Tiago Tresoldi (tresoldi@shh.mpg.de).
The author has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No. ERC Grant #715618, Computer-Assisted Language Comparison.
If you use distfeat
or the standard feature model distributed with it,
please cite it as:
Tresoldi, Tiago (2020). DistFeat, a Python library for manipulating segmental and distinctive features. Version 0.1.1. Jena. DOI: 10.5281/zenodo.3902005
In BibTeX:
@misc{Tresoldi2020distfeat,
author = {Tresoldi, Tiago},
title = {DistFeat, a Python library for manipulating segmental and distinctive features. Version 0.1.},
howpublished = {\url{https://github.com/tresoldi/distfeat}},
address = {Jena},
year = {2020},
doi = {10.5281/zenodo.3902005}
}
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
File details
Details for the file distfeat-0.2.tar.gz
.
File metadata
- Download URL: distfeat-0.2.tar.gz
- Upload date:
- Size: 112.6 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/3.1.1 pkginfo/1.5.0.1 requests/2.23.0 setuptools/41.2.0 requests-toolbelt/0.9.1 tqdm/4.46.1 CPython/3.7.7
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | e88b880be2bc8f1eec8df821a28c139121a675b104b9afa968056178c659e000 |
|
MD5 | 716a18e01d7d4a4e457c8a67d88c9f80 |
|
BLAKE2b-256 | 9ba35a0dd1f9bd683cafa0567e26a8f86651bc079a65b22acd3a7194172e979d |
File details
Details for the file distfeat-0.2-py3-none-any.whl
.
File metadata
- Download URL: distfeat-0.2-py3-none-any.whl
- Upload date:
- Size: 140.9 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/3.1.1 pkginfo/1.5.0.1 requests/2.23.0 setuptools/41.2.0 requests-toolbelt/0.9.1 tqdm/4.46.1 CPython/3.7.7
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | e1eea4f5ad42bd13061b04a31b471e26c265871ee3e5f59f3745a00980abd05a |
|
MD5 | 69b75d6728998928d98317259b26bd13 |
|
BLAKE2b-256 | b649c3be0127bb2f77e39b57b4422461a1c4609c9fc43c651d0cb6b100c48c5f |