Uzbek Morphological Analyzer using Complete Set of Ending (CSE)

Project Description

uzmorph is a professional-grade morphological analyzer for the Uzbek language that combines a massive lexicon (~122k stems) with CSE (Complete Set of Endings) morphological rules. It supports robust suffix stripping and multi-POS disambiguation for high-accuracy linguistic analysis.

You can try the live demo here: Hugging Face Space

Key Features

• High Accuracy: Achieved 93% Word Coverage Accuracy on a sample of 20K unique Uzbek words.
• Massive Lexicon: Built with over 122K unique stem-POS pairs.
• Rule-Based CSE Engine: Implements the Complete Set of Endings paradigm for agglutinative suffix analysis.
• Multi-POS Support: Handles ambiguous words (e.g., ot as both Noun "horse" and Verb "throw") by validating suffix rules against lexicon POS.
• Rich Morphological Tagging: Extracts detailed features including part-of-speech (POS), tense, person, possession, case, and voice.
• Flat JSON Output: Returns analysis results in a developer-friendly, flattened JSON-compatible format.

Installation

pip install uzmorph

Quick Start

from uzmorph import UzMorph

# Initialize the analyzer
analyzer = UzMorph()

# Analyze a word
results = analyzer.analyze("maktabimda")

# Formatted console print
analyzer.print_result(results)

JSON Result Sample

Each analysis result is a dictionary containing the following structure:

[
    {
        "word": "maktabimda",
        "stem": "maktab",
        "lemma": "maktab",
        "cse": "imda",
        "cse_formula": "(i)mda",
        "pos": "NOUN",
        "possession": "1",
        "cases": "Locative",
        "singular": "1",
        "syntactical_affixes": "(i)m da",
        "note": null,
        "ball": 308
    }
]

API Reference

UzMorph Class

• analyze(word, pos_filter=None): Performs morphological analysis and returns a list of results.
• print_result(results): Prints formatted output to the console.
• get_pos_list(): Returns a formatted string of all available POS tags.
• get_features_list(): Returns a list of all possible property keys in the result.

Project Structure (for PyPI)

For a clean PyPI contribution, only the following files are included in the package:

• uzmorph/ - Core package directory containing implementation and data files (root.csv, cse.csv, etc.)
• pyproject.toml - Package metadata and build configuration.
• LICENSE - MIT License file.
• README.md - Documentation.
• MANIFEST.in - File inclusion rules for the source distribution.

License

MIT