verbe-conjugaison-académie-française

A comprehensive dataset of 6,298 French verbs with complete conjugation data extracted from the Académie française's official dictionary. This repository includes both the parsed conjugation data and the Python scripts used to generate it.

Dataset Overview

• 6,298 verbs from the 9th edition of the Académie française dictionary
• Complete conjugation tables across all moods, tenses, and persons
• Gender-aware conjugations — separate masculine/feminine forms for all persons when they differ (passive voice, être auxiliaries, pronominal compounds)
• Three output formats: JSON (formatted), JSON (minified), and SQLite3 database
• Multi-threaded parser for efficient data generation
• 1990 orthography reform support with variant tracking

Quick Start

Installation

# Clone the repository
git clone https://github.com/ShingZhanho/verbe-conjugaison-academie-francaise.git
cd verbe-conjugaison-academie-francaise

# Install dependencies
pip install -r requirements.txt

Usage

# Generate JSON files (uses 4 threads by default)
python -m verbe_af

# Generate JSON + SQLite database
python -m verbe_af --gen-sqlite3

# Generate with 8 threads for faster processing
python -m verbe_af --max-threads 8

# Force fresh data (ignore cache)
python -m verbe_af --ignore-cache

# Generate infinitives list only
python -m verbe_af --gen-infinitives

# Verbose output for debugging
python -m verbe_af --verbose

# Write log to file (in addition to terminal)
python -m verbe_af --log-file crawl.log

Output Files

JSON Files (verbs.json and verbs.min.json)

Two JSON files are generated:

• verbs.json - Human-readable formatted conjugation data
• verbs.min.json - Minified version for production use

JSON Structure

{
  "aller": {
    "h_aspire": false,
    "rectification_1990": false,
    "rectification_1990_variante": null,
    "voix_active_etre": {
      "participe": {
        "present": "allant",
        "passe": {
          "sm": "allé",
          "sf": "allée",
          "pm": "allés",
          "pf": "allées",
          "compound_sm": "étant allé",
          "compound_sf": "étant allée",
          "compound_pm": "étant allés",
          "compound_pf": "étant allées"
        }
      },
      "indicatif": {
        "present": {
          "1sm;1sf": "vais",
          "2sm;2sf": "vas",
          "3sm;3sf": "va",
          "1pm;1pf": "allons",
          "2pm;2pf": "allez",
          "3pm;3pf": "vont"
        },
        "passe_compose": {
          "1sm": "suis allé",
          "1sf": "suis allée",
          "2sm": "es allé",
          "2sf": "es allée",
          "3sm": "est allé",
          "3sf": "est allée",
          "1pm": "sommes allés",
          "1pf": "sommes allées",
          "2pm": "êtes allés",
          "2pf": "êtes allées",
          "3pm": "sont allés",
          "3pf": "sont allées"
        }
      }
    }
  }
}

Key Features

Voices (voix):

• voix_active_avoir - Active voice with auxiliary "avoir"
• voix_active_etre - Active voice with auxiliary "être"
• voix_active - Active voice (defective verbs with unknown auxiliary)
• voix_passive - Passive voice
• voix_prono - Reflexive/pronominal form

Moods (modes):

• participe - Participle (present and past)
• indicatif - Indicative (8 tenses)
• subjonctif - Subjunctive (4 tenses)
• conditionnel - Conditional (2 tenses)
• imperatif - Imperative (2 tenses)

Persons (personnes) — full {1,2,3} × {s,p} × {m,f,n} grid:

• 1sm / 1sf — je (first person singular masculine / feminine)
• 2sm / 2sf — tu (second person singular masculine / feminine)
• 3sm — il (third person singular masculine)
• 3sf — elle (third person singular feminine)
• 3sn — on (third person singular neutral — used by reciprocal verbs like s'entraider)
• 1pm / 1pf — nous (first person plural masculine / feminine)
• 2pm / 2pf — vous (second person plural masculine / feminine)
• 3pm — ils (third person plural masculine)
• 3pf — elles (third person plural feminine)

Key merging — when multiple person keys share the same conjugation value within a tense, they are merged into a single semicolon-separated key:

• Simple tenses (no participle agreement): "1sm;1sf", "2sm;2sf", "3sm;3sf", "1pm;1pf", "2pm;2pf", "3pm;3pf" — 6 keys
• Compound tenses with être (participle agrees in gender): all 12 keys separate
• Passive voice simple tenses: keys merge across persons with the same auxiliary and participle form (e.g. "1sm;2sm" when both use étais combiné)

SQLite3 Database (verbs.db)

A normalized relational database (~361 MB) with ~1.6M conjugation rows optimized for queries. The personne column uses the same merged key format as the JSON output (e.g. "1sm;1sf").

Database Schema

verbes table - Core verb metadata

CREATE TABLE verbes (
    id INTEGER PRIMARY KEY AUTOINCREMENT,
    infinitif TEXT UNIQUE NOT NULL,
    h_aspire BOOLEAN NOT NULL,
    rectification_1990 BOOLEAN NOT NULL,
    rectification_1990_variante TEXT
);

conjugaisons table - Person conjugations

CREATE TABLE conjugaisons (
    id INTEGER PRIMARY KEY AUTOINCREMENT,
    verbe_id INTEGER NOT NULL,
    voix TEXT NOT NULL,
    mode TEXT NOT NULL,
    temps TEXT NOT NULL,
    personne TEXT NOT NULL,
    conjugaison TEXT NOT NULL,
    FOREIGN KEY (verbe_id) REFERENCES verbes(id) ON DELETE CASCADE,
    UNIQUE(verbe_id, voix, mode, temps, personne)
);

participes table - Participle forms

CREATE TABLE participes (
    id INTEGER PRIMARY KEY AUTOINCREMENT,
    verbe_id INTEGER NOT NULL,
    voix TEXT NOT NULL,
    forme TEXT NOT NULL,
    participe TEXT NOT NULL,
    FOREIGN KEY (verbe_id) REFERENCES verbes(id) ON DELETE CASCADE,
    UNIQUE(verbe_id, voix, forme)
);

Optimized Indexes

CREATE INDEX idx_verbes_infinitif ON verbes(infinitif);
CREATE INDEX idx_verbes_variantes ON verbes(rectification_1990_variante);
CREATE INDEX idx_conjugaisons_recherche ON conjugaisons(verbe_id, voix, mode, temps, personne);
CREATE INDEX idx_conjugaisons_texte ON conjugaisons(conjugaison);
CREATE INDEX idx_participes_recherche ON participes(verbe_id, voix, forme);

Example Queries

-- Get all present indicative conjugations for "être"
SELECT personne, conjugaison
FROM conjugaisons c
JOIN verbes v ON c.verbe_id = v.id
WHERE v.infinitif = 'être' 
  AND voix = 'voix_active_avoir'
  AND mode = 'indicatif' 
  AND temps = 'present'
ORDER BY personne;

-- Find verbs with 1990 orthography reform variants
SELECT infinitif, rectification_1990_variante
FROM verbes
WHERE rectification_1990 = 1;

-- Search for all conjugations containing "aient"
SELECT v.infinitif, c.voix, c.mode, c.temps, c.personne, c.conjugaison
FROM conjugaisons c
JOIN verbes v ON c.verbe_id = v.id
WHERE c.conjugaison LIKE '%aient%'
LIMIT 20;

-- Get all participle forms for "aller"
SELECT voix, forme, participe
FROM participes p
JOIN verbes v ON p.verbe_id = v.id
WHERE v.infinitif = 'aller';

-- Count conjugations by mood
SELECT mode, COUNT(*) as total
FROM conjugaisons
GROUP BY mode
ORDER BY total DESC;

Command Line Options

Basic Options

Option	Description	Default
--verbose, -v	Enable detailed logging	False
--ignore-cache	Force fresh data fetch (ignore cached HTML and parsed store)	False
--gen-sqlite3	Generate SQLite database file	False
--gen-infinitives	Generate infinitives list only	False
--log-file PATH	Write log to file (in addition to terminal)	—

Performance Options

Option	Description	Default
--max-threads N	Number of concurrent parsing threads	4
--max-retry N	Maximum HTTP request retries	5
--requests-delay MS	Delay between requests (milliseconds)	500

Advanced Options

Option	Description
--user-agent AGENT	Custom user agent string
--jsession-id ID	Override JSESSION_ID cookie

Examples

# Fast generation with 8 threads
python -m verbe_af --max-threads 8 --gen-sqlite3

# Conservative mode (slower, but safer for rate limiting)
python -m verbe_af --max-threads 2 --requests-delay 1000

# Debug mode with verbose output
python -m verbe_af --verbose --max-threads 1

# Generate fresh database ignoring cache
python -m verbe_af --ignore-cache --gen-sqlite3 --max-threads 8

Data Accuracy

Gender Agreement

The parser correctly extracts gender-specific conjugations from the Académie française dictionary for all persons:

• All persons receive distinct masculine/feminine forms when the verb includes a gender-agreeing past participle (passive voice, être auxiliary compound tenses, pronominal compound tenses)
• Example: "je suis allé" (1sm) vs "je suis allée" (1sf)
• Example: "il est allé" (3sm) vs "elle est allée" (3sf)
• Example: "ils sont allés" (3pm) vs "elles sont allées" (3pf)
• When masculine and feminine forms are identical (simple tenses, avoir auxiliary), person keys are merged: "1sm;1sf": "vais"
• Reciprocal verbs using the pronoun on are assigned the neutral key 3sn (e.g., s'entraider: "3sn": "s'entraide")

1990 Orthography Reform

Verbs with reformed spellings are tracked:

• Both variants appear as separate entries (e.g., "connaître" and "connaitre")
• Metadata fields: rectification_1990 (boolean) and rectification_1990_variante (string)

🔧 Technical Details

Requirements

• Python: 3.13+ (tested, lower versions may work)
• Dependencies: BeautifulSoup4 (lxml), requests
• Platform: Cross-platform (Windows, macOS, Linux)

Architecture

The codebase is organised as a Python package (verbe_af/) with clean separation of concerns:

Module	Responsibility
verbe_af/cli.py	argparse CLI and main dispatch
verbe_af/config.py	Injectable Config dataclass (replaces mutable globals)
verbe_af/client.py	DictionaryClient — requests.Session-based HTTP
verbe_af/parser.py	Conjugation HTML → structured dict
verbe_af/transformer.py	Normalise parsed data, 1990 reform handling
verbe_af/cache.py	HTML cache helpers, ParsedStore (SQLite KV), JSON merge
verbe_af/crawler.py	VerbCrawler — threaded orchestration
verbe_af/constants.py	Immutable constants, VoiceType enum, person key maps
verbe_af/exceptions.py	CrawlerError hierarchy
verbe_af/extensions/	Optional generators (infinitives, SQLite)
audit/	PySide6 GUI for manual data verification

Audit GUI

A PySide6-based manual verification tool for comparing parsed JSON data against source HTML:

pip install -r audit/requirements.txt
python -m audit --auditor yourname

Features: split-pane layout (JSON vs source HTML), per-tense review units, per-form flag buttons, status/voice filters, CSV export of flagged items, JSONL progress file (Git-collaborative), keyboard shortcuts (Alt+O/Alt+F/Alt+S/Alt+N/Alt+P).

Related Resources

• Académie française Dictionary
• French Orthography Reform (1990)