A French-English text realizer
Project description
pyrealb - A Python Bilingual Text Realizer
Version 3.1 - September 2024
pyrealb is a Python adaptation of the JavaScript jsRealB text realizer with the same constituent and dependency syntax notation. It facilitates its integration within Python applications by simply adding
from pyrealb import *
Version 3.0.0 was a major code reorganization, but without any new feature, to clearly separate language dependent parts from the language independent ones. This organization is described here .
The use of pyrealb for Bilingual Data-to-text generation is described in this document.
Installing the distribution package from PyPI
pip install pyrealb
Caution: do not forget the b
at the end of pyrealb
. On PyPI, there is an unrelated package pyreal
for evaluating and deploying human readable machine learning explanations.
Upgrading the version
pip install pyrealb --upgrade
Building and installing the package from the sources
cd
into this directory (withpyproject.toml
file)- If appropriate, change value of
version
insetup.cfg
to match the value ofpyrealb_version
inutils.py
- Build the distribution package
python3 -m build
- Install with
python3 -m pip install .
First realization tests at the Python 3 prompt
from pyrealb import *
loadEn()
print(S(Pro("I").g("f"),VP(V("say"),"hello",PP(P("to"),NP(D("the"),N("world"))))))
- this should print
She says hello to the world.
print(root(V("say").t("ps"),subj(Pro("him").c("nom")),comp(N("goodbye"))).typ({"neg":True}))
- this should print
He did not say goodbye.
Use pyrealb in a Jupyter notebook thru Binder
- in a browser, load one of these links:
Directories
src/pyrealb
__init__.py
: import classes and functions and export relevant symbols.Constituent.py
: Constituent is the top class for methods shared between Phrases and TerminalsConstituentEn.py
,ConstituentFr.py
: English and French specific processing ofConstituent
Dependent.py
: subclass of Constituent for creating complex phrases using dependenciesDependentEn.py
,DependentFr.py
: English and French specific processing ofDependent
Lexicon.py
: class to access lexicon entries and syntactic rulesLICENSE.txt
: Apache 2.0 LicenseNonTerminalEn.py
,NonTerminalFr.py
: language dependent processing common toPhrase
andDependent
Number.py
: utility functions for dealing with number formattingPhrase.py
: subclass of Constituent for creating complex phrasesPhraseEn.py
,PhraseFr.py
: English and French specific processing ofPhrase
Terminal.py
: subclass of Constituent for creating a single unit (most often a single word)TerminalEn.py
,TerminalFr.py
: English and French specific processing ofTerminal
utils.py
: some useful functions
./src/pyrealb/data
:LICENSE.txt
: Creative Common licenselexicon-en.json
: English lexicon (33,932 entries) in json formatrule-en.js
: English conjugation and declension tableslexicon-fr.json
: French lexicon (52,547 entries) in json formatrule-fr.js
: French conjugation and declension tables
Nota bene:
- In the following directories, the
__init__.py
file is used to set the appropriate search path for pyrealb functions; this ensures that the current Python source files are used for execution. - Some directories include
markup.py
which should be loaded usingpip
. Unfortunately I never managed to make this "piped" version work, it does not import the nameoneliner
although it should. It works only if the file is in the local directory.
-
docs
: in both English and French.documentation.html
: generated documentation (used for consultation) DO NOT EDIT directly Online versiondocumentation.py
: Python program for generatingdocumentation.html
usingmarkup.py
once this is rundocumentation.html
should be copied athttp://www.iro.umontreal.ca/~lapalme/pyrealb/documentation.html
which is used for consultationstyle.css
: style sheet for the documentationuserinfos.py
: definitions of variables containing the examplesuser.js
: Python helper script.
-
IDE
: Integrated Development Environmentide.py
: built on the Python read-eval-print loop, it imports pyrealb to get the realization of an expression, to consult the lexicon, the conjugation and declension tables. It is also possible to get a lemmatization: i.e. the pyrealb expression corresponding to a form.README.html
: documentation and examples
-
tests
: unit tests of special features of pyrealb in both French and English. Files have the pattern*_{en|fr}.py
test.py
: simplistic function to check if a function returns the expected answer and display appropriate messagetestAll.html
: run this file to run all tests
Demos
99bottlesofbeer/99bottlesofbeer.py
: simple generation of a classic repetitive text in English.basketball/sportsettsum.py
: generation of French and English basketball summaries paper describing the approachBilinguo/bilinguo.py
: generation of translation drill exercisesdev_example/dev_example.py
: examples of English and French expressions to be realized and checked against expected output,
useful for debugging when adding a new expression and enabling tracingevenements/evenements.py
: Description (in French) of a list of events, it creates HTML.flight_infos/README.md
: development of a RASA NLG server giving information about flights, aircrafts, etc...gen_from_words.py
: generation of English and French sentences from a plain list of words, adding some structure.gophypi/amr2text.py
: generate a literal reading of an AMR (Abstract Meaning Representation); paper describing the approachinflection/inflection.py
: French or English conjugation and declension of a form.kilometresapied/kilometresapied.py
: simple generation of a classic repetitive text in French.methodius/methodius.py
: generation of English sentences from a logical form expressed in XML.randomgen/randomgen.py
: Generation of random English sentencesRDFpyrealb/WebGenerate.py
: Generation from RDF triplesreport/report.py
: Single sentence parameterized by language, tense and subject using two different program organizationvariantes/variantes.py
: French or English sentences realized with all possible sentence modifiers; some challenging examples are inexamples.py
.weather/Bulletin.py
: French and English weather bulletins generated from information in a json-line file. (weather-data.jsonl
). It uses the packages in theRealization
directory.
Licenses
- pyrealb source code is licensed under Apache-2.0
- linguistic resources in the
./data
directory are licensed under CC-BY-SA-4.0
Contact
Acknowledgement
Thanks to Fabrizio Gotti, François Lareau and Ludan Stoeckle for interesting suggestions.
For the maintainer mainly
Updating package version on PyPI
see this tutorial
These steps take for granted that the password for PyPI has already been given...
- Update version number in
setup.cfg
(it should be the same aspython_version
insrc/pyrealb/utils.py
and at the beginning of this document). - Run
docs/documentation.py
to update the version number indocs/documentation.html
cd
into the directory with thepyproject.toml
file (the same as thisREADME.md
)- Build the distribution package
python3 -m build
- Upload to PyPi the last version I.J.K
twine upload dist/*-I.J.K.*
- Install new version from PyPI
python3 -m pip install pyrealb --upgrade
Useful trick for debugging with breaking point and tracing in PyCharm
- add
pyrealb
expression to debug at the end ofdemo/dev_example/dev_example.py
- comment the line calling
testPreviousExamples()
- debug
demo/dev_example/dev_example.py
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.