Python library for verb conjugation in French, Spanish, Catalan, Italian, Portuguese, and Romanian, enhanced by machine learning

verbecc python library

[EN] Verbs completely conjugated: verb conjugations for French, Spanish, Portuguese, Italian, Romanian and Catalan, enhanced by machine learning

[CA] Verbs completament conjugats: conjugacions verbals per a francès, espanyol, portuguès, italià, romanès i català, millorades per l'aprenentatge automàtic

[ES] Verbos completamente conjugados: conjugaciones de verbos en francés, español, portugués, italiano, rumano y catalán, mejoradas por aprendizaje automático

[FR] Verbes complètement conjugués: conjugaisons des verbes français, espagnol, portugais, italien, roumain et catalan, à l'aide de l'apprentissage automatique

[IT] Verbi completamente coniugati: coniugazioni di verbi per francese, spagnolo, portoghese, italiano, rumeno e catalano, migliorate dall'apprendimento automatico

[PT] Verbos completamente conjugados: conjugações verbais para francês, espanhol, português, italiano, romeno e catalão, aprimoradas pelo aprendizado de máquina

[RO] Verbe complet conjugate: conjugări de verbe pentru franceză, spaniolă, portugheză, italiană, română și catalană, îmbunătățite de învățarea automată

Quick Start
Live Demo
Example Output
What's new in Verbecc 2.0
Academic publications referencing Verbecc
Typing - Parameter and Data Type Annotations
Multi-Language Conjugation
Multi-Language Conjugation using English mood and tense names via localization module
Credits

Live Demo

Example Output

| Français / French | Català / Catalan | Español / Castellano / Spanish | Português / Portuguese | Italiano / Italian | Română / Romanian | | ------ | ------ | ------- | -------- | --------- | ------ | | Français / French | Català / Catalan | Español / Castellano /Spanish | Português / Portuguese | Italiano / Italian | Română / Romanian | | French être (to be) | Catalan ser (to be) | Spanish ser (to be) | Portuguese ser (to be) | Italian essere (to be) | Romanian fi (to be) | | French se lever (to lift oneself) | | | | | | | French ubériser (to "uberize") (unknown verb conjugated with ML template prediction)) | | | | | |

Features

Multilingual
- Conjugate verbs in six romance languages: French, Spanish, Portuguese, Italian, Romanian, Catalan
- Includes Spanish voseo conjugation, with regional options in development.
- Predict conjugation of unknown verbs with 99% accuracy using machine learning techniques
- Conjugate thousands of known verbs without machine learning, using simple string transformations based on XML conjugation templates
Complete
- Includes both simple and compound conjugations (i.e. with helping/auxiliary verbs)
- Includes alternate conjugations (for regional variations, e.g. Catalan vs. Valencian)
- Includes inflections for all genders where applicable
- Includes inlections for misc. pronouns such as the Spanish pronouns usted and ustedes and the French pronoun on.
Quality
- Fully type-annotated python library
  - Unit-tests require type-annotations on everything
- Typed return data
- Meticulously organized source tree
- Has a plethora of unit-tests to ensure correctness of verb conjugations
- Continuous Integration with GitHub Actions CI/CD pipeline
  - CI tests python 3.9, 3.10, 3.11, 3.12, 3.13 and 3.14.
- Dependencies: scikit-learn, scipy, numpy, lxml, pyaml, jsbeautifier, importlib_resources
Trusted
- Cited in academic publications

Quick Start

git clone https://github.com/bretttolbert/verbecc.git
cd verbecc
pip install .

Academic publications referencing verbecc

Segura Lores, Alba. (2025) Estudio de la variación del sujeto pronominal en la ciudad de Málaga: sociolingüística cognitiva de su producción y percepción. Universität Heidelberg. pp. 69. [DOI:10.11588/heidok.00037508]

What's new in Verbecc 2.0

| verbecc 1.x | verbecc 2.x | | --- | --- | | lang='fr' | lang=Lang.fr / from verbecc import LangCodeISO639_1 as Lang | | mood="indicatif" | mood=Moods.fr.Indicatif / from verbecc import Moods | | tense="présent" | tense=Tenses.fr.Présent / from verbecc import Tenses | | gender='f' | gender=Gender.f / from verbecc import Gender | | person="1s" | person=Person.First, number=Number.Singular / from verbecc import Person, Number | | Conjugations include masculine pronouns (default) or feminine but not both | All pronouns, including both masculine and feminine third-person pronouns are included | | lang_specific_options is a parameter of the conjugate method | lang_specific_options is a parameter of the CompleteConjugator class constructor | | gender is a parameter of the conjugate method | there is no gender parameter, instead all possible gender inflections are returned | | alternate_options is a parameter of the conjugate method | there is no alternate_options parameter, instead all possible conjugations, including alternates, are returned (use c[0] to get default conjugation, c[1] to get first alternate, etc.) | | Spanish Conjugations include tú (default) or vos but not both | All pronouns, including both tú and vos are included | | Pronouns such as French on and Spanish usted/ustedes not included | French on and Spanish usted/ustedes pronouns are included | | Array index is used to determine Person, i.e. 1s, 2s, 3s, 1p, 2p, 3p | Each Conjugation object in the TenseConjugation has Person, Number and Gender values (any of which may be None if not-applicable) | | Returned objects are primitive (Dict) data types | Returned wrapper objects are subclasses of AbstractConjugation (e.g. CompleteConjugation) with get_data() and to_json() methods | | Conjugator returns CompleteConjugationData | CompleteConjugator returns wrapper type CompleteConjugation, CompleteConjugation.get_data() returns CompleteConjugationData | | (no wrapper types) | Wrapper types hierarchy: CompleteConjugation > MoodsConjugation > MoodConjugation > TenseConjugation > Conjugation -> conjugations: List[str] | | Primitive data types hierarchy: Conjugation > MoodsConjugation > MoodConjugation > TenseConjugation > PersonConjugation | Primitive data types hierarchy: CompleteConjugationData > MoodsConjugationData > MoodConjugationData > TenseConjugationData > ConjugationData -> conjugations: List[str] | | pred_score was always included in the output | pred_score is only included in output if predicted is true | | Only returned primitive Python data | Conjugation objects have both .to_json() and .to_yaml() methods | |

Typing - Parameter and Data Type Annotations

Originally verbecc used strings for most parameters. verbecc is now fully type-annotated but strings are still supported for backwards-compatibility and ease of use. This is accomplished using StrEnum for parameters and by defining a hierarchy of typing type definitions for the returned data objects (See conjugation.py).

E.g.:

>>> from verbecc import grammar_defines, localization, Moods, Tenses, Person, Number, Gender, LangCodeISO639_1 as Lang
>>> xmood = localization.xmood
>>> xtense = localization.xtense
>>> grammar_defines.SUPPORTED_LANGUAGES[Lang.fr]
'français'
>>> xtense(Lang.fr, Tenses.en.Present)
<TenseFr.Présent: 'présent'>
>>> xmood(Lang.fr, Moods.en.Subjunctive)
<MoodFr.Subjonctif: 'subjonctif'>
>>> Gender.f
<Gender.f: 'f'>
>>> Number.Singular
<Number.Singular: 's'>
>>> Person.First
<Person.First: '1'>

Multi-Language Conjugation

>>> from functools import partial
>>> from verbecc import CompleteConjugator, LangCodeISO639_1 as Lang, grammar_defines, Moods, Tenses
>>> ccgs = {lang : CompleteConjugator(lang) for lang in grammar_defines.SUPPORTED_LANGUAGES}

>>> print([c[0] for c in ccgs[Lang.fr].conjugate('être')[Moods.fr.Indicatif][Tenses.fr.Présent]])
['je suis', 'tu es', 'il est', 'elle est', 'on est', 'nous sommes', 'vous êtes', 'ils sont', 'elles sont']
>>> print([c[0] for c in ccgs[Lang.es].conjugate('ser')[Moods.es.Indicativo][Tenses.es.Presente]])
[

Verbecc

Install / Use

README