Lexika is a Python desktop application built around an offline French dictionary of nearly 900,000 entries. It offers four integrated tools:
- Dictionary
- Lexicon
- Quiz
- Analyzer
-
Dictionary: nearly 900,000 French entries, including conjugated verb forms
-
Lexicon: personal lexicon to save words you want to remember
-
Quiz: vocabulary quiz to review the words in your lexicon
-
Analyzer: Word analyzer to filter the dictionary with multiple cumulative criteria
- Python 3.10 or higher
- Dependencies listed in
requirements.txt
pip install -r requirements.txtrequirements.txt:
customtkinter
pillow
git clone https://github.com/Kartmaan/lexika-fr.git
cd lexika-frpip install -r requirements.txtpython main.pyOn first launch, if data/french_dict.db is missing, a setup window appears automatically and offers two options:
- Download the dictionary from Hugging Face (~280 MB)
- Import a compatible
.dbfile already on your disk
The file is automatically validated before use (extension, SQLite structure, data presence).
lexika-fr/
├── main.py # Entry point
├── requirements.txt
├── assets/
│ ├── icon.png # Linux icon
│ ├── icon.ico # Windows icon
│ └── icon.icns # macOS icon
├── core/
│ ├── config.py # Global visual identity and grammatical mapping
│ ├── dictionary.py # SQLite queries, suggestions, analyzer
│ └── lexicon.py # Lexicon JSON management
├── ui/
│ ├── app.py # Main window and tabs
│ ├── setup_window.py # First-launch setup window
│ ├── tab_dictionary.py # Dictionary tab
│ ├── tab_lexicon.py # Lexicon tab
│ ├── tab_quiz.py # Quiz tab
│ └── tab_analyzer.py # Analyzer tab
└── data/
├── french_dict.db # SQLite database (generated at setup)
└── lexicon.json # Personal lexicon (auto-created)
The main tab of the application. Allows access to the definitions of a word
Search
- Type a word in the search field and confirm with the button or the
Enterkey - Search is case-insensitive
Results display
- Definitions are grouped by part of speech (Noun, Verb, Adjective...) with a color badge
- Each definition is numbered and may include:
- Gender if the word is an noun
- Hierarchical sub-definitions
- Usage examples in italics
- Register tags (familiar), semantic tags [figurative] or domain tags ‹music›
Word not found
- If the word does not exist in the dictionary, Lexika automatically suggests similar words
- The fuzzy search handles missing accents: typing
elementsuggestsélément, typingenchevetresuggestsenchevêtré - Clicking a suggestion directly loads its definition
Copy to clipboard
- Copies the selected word and its definitions to the clipboard.
Add to lexicon
- An Add to lexicon button is available below each result
- If the word is already in the lexicon, a message notifies you
Personal lexicon of words added from the dictionary.
Left column - word list
- Saved words appear in alphabetical order as clickable tiles
- Words sourced from the dictionary appear in blue
- Custom words appear in purple
- A search field allows you to filter the list by typing part of a word
Right column - definitions
- Clicking a tile immediately displays the full definition in the right column
- A View in dictionary button navigates to the Dictionary tab to show the original entry (available only for dictionary-sourced words)
Lexicon management
- Remove a word from the lexicon using the dedicated button
- Add a custom word: opens a form to enter a word and one or more free-form definitions - useful for technical terms, jargon, or neologisms absent from the dictionary
- Export the lexicon to a
.jsonfile of your choice - Import a previously exported lexicon (existing words are preserved and new ones are merged in)
A tool for reviewing the vocabulary saved in your lexicon.
Word side
Definition side
How it works
- The quiz can only start if the lexicon contains at least one word
- Words are drawn in a random order at the start of each session
- Each word appears only once per session
The flashcard
- The card first displays the word to define on a blue background
- The See the answer button flips the card: it turns green and reveals the definition(s)
- The See the word button flips it back to the word side
- The Next word button moves to the next word in the session
End of session
- When all words have been reviewed, a completion screen shows the number of words covered
- A Play again button starts a new session in a different random order
A word-filtering tool that queries the full dictionary using multiple cumulative criteria.
Available filters
All filters are optional and combinable. The more filters are active, the more precise the results.
| Filter | Description | Example input |
|---|---|---|
| Length | Exact number of letters | 7 |
| Starts with | The word must begin with this prefix | gr |
| Ends with | The word must end with this suffix | it |
| Contains | Letters the word must include (continuous or space-separated) | au or a u |
| Not contain | Letters the word must not include | bx or b x |
| Anagram | The word must be an exact anagram of these letters | carte or c a r t e |
| Letter at position | One or more positional constraints (1-indexed) | Pos 2 = r, Pos 4 = t |
| Exclude compound words | Removes hyphenated and multi-word entries (on by default) | toggle |
Combining filters
Filters are applied as cascading SQL conditions, each active filter narrows down the previous results. For example:
Length = 7, Starts with = g, Ends with = it,
Contains = au, Not contain = b,
Letter at position: Pos 2 = r, Pos 4 = t
→ Gratuit, Grutait
Anagram search
Perfect anagrams
The anagram filter finds all words in the dictionary that use exactly the same letters as the input, regardless of order. Accented variants are handled automatically, searching carte will find Carte, Carté, Trace, Tracé, Acter, Caret, and more.
Partial anagrams
When enabled, the anagram search no longer requires words to use all the provided letters, it instead returns any word that can be formed using a subset of them. For example, searching for "carte" with partial anagrams enabled, the results are expanded to include shorter words such as "Car", "Arc", "Rat" or "Acte", as long as each letter used appears in the original pool and isn't used more times than it's available. Words shorter than three letters are excluded from partial results. This mode is particularly useful for board game players, such as Scrabble, who may not always be able to place all their tiles at once.
Anagram can be combined with other filters: for instance, adding Starts with = t to
carterestricts results to anagrams beginning with 't' (Trace,Tracé).
Letter at position
Click + Add position to add a positional constraint row (position + letter). Multiple rows can be stacked for finer control. Each row can be removed independently.
Results
- Results are displayed as clickable tiles, sorted alphabetically
- Up to 500 words are shown per search; a notice appears if results are truncated
- Clicking a tile navigates directly to the Dictionary tab to display the word's full definition
The dictionary is stored as a SQLite database (.db) rather than a flat file format such as CSV, JSON, or Parquet. This choice was driven by the specific requirements of a desktop dictionary application handling nearly 900,000 entries.
Instant lookups via B-tree indexing
SQLite maintains a B-tree index on the forme column, meaning an exact word search across 900,000 entries completes in under a millisecond, regardless of the position of the word in the alphabet. A flat file would require a full scan on every keystroke.
Cascading filters without loading data into memory
The Analyzer tab runs multi-criteria queries (length, prefix, suffix, contained letters, positional constraints) entirely within SQLite's query engine. Each active filter is translated into a SQL WHERE clause, and only the matching rows are ever read from disk. Equivalent operations on a Pandas DataFrame loaded from a Parquet or CSV file would iterate over all rows in Python, taking several seconds per query on this dataset size.
Zero memory footprint at startup
SQLite is a lazy reader: the application connects to the file but loads nothing into RAM until a query is made. The 280 MB database occupies virtually no memory at rest. A JSON or Parquet approach would require loading the entire dataset into memory on startup, adding 20–270 MB to the application's footprint depending on the format.
Lexika's graphical interface is built with CustomTkinter, a modern UI library that extends Python's standard tkinter with a dark-mode-first design system, rounded widgets, and a theming engine, without requiring any external graphical toolkit such as Qt or GTK.
No system dependencies
Standard tkinter ships with Python on all major platforms. CustomTkinter adds only a pure-Python layer on top of it, meaning the entire UI stack installs with a single pip install customtkinter. There is no native library to compile, no system package to install, and no compatibility issue between platforms.
Cross-platform out of the box
The same code runs on Linux, Windows, and macOS without conditional branches or platform-specific backends. Window management, icon handling, and font rendering adapt automatically to the host OS.
Lightweight footprint
Unlike Electron-based or Qt-based alternatives, CustomTkinter adds negligible overhead to startup time and RAM usage. For an offline dictionary application where the database itself weighs 280 MB, keeping the UI layer minimal was a deliberate choice.
The dictionary is derived from WiktionaryX, a XML structured lexical resource parsed from the French Wiktionary, produced by Franck Sajous, CNRS research engineer and lecturer in Language Sciences at the University of Toulouse.
Original source: http://redac.univ-tlse2.fr/lexiques/wiktionaryx.html
The french_dict.db file is hosted separately on Hugging Face (CC BY-SA 4.0 license): https://huggingface.co/datasets/Kartmaan/french-dictionary
The Hugging Face repository also contains this same database in
.parquetformat, which is more useful for machine learning and data science projects.
| Component | License |
|---|---|
| Source code (this repository) | MIT |
french_dict.db database |
CC BY-SA 4.0 (derived from Wiktionary) |