Skip to content

Latest commit

 

History

History
55 lines (35 loc) · 2.07 KB

README.md

File metadata and controls

55 lines (35 loc) · 2.07 KB
Raw

🧩 TrTokenizer

A simple sentence tokenizer.

Python Version PyPI Version

Overview

TrTokenizer is a comprehensive solution for Turkish sentence and word tokenization, tailored to accommodate extensive language conventions. If you're seeking robust, fast, and accurate tokenization for natural language models, you've come to the right place. Our sentence tokenization approach employs a list of non-prefix keywords found in the 'tr_non_suffixes' file. Developers can conveniently expand this file, and lines starting with '#' are treated as comments. We've designed regular expressions that are pre-compiled for optimal performance.

Installation

You can install TrTokenizer via pip:

pip install trtokenizer

Usage

Here's how you can use TrTokenizer in your Python projects:

from trtokenizer.tr_tokenizer import SentenceTokenizer, WordTokenizer

# Initialize a SentenceTokenizer object
sentence_tokenizer = SentenceTokenizer()

# Tokenize a given paragraph as a string
sentence_tokenizer.tokenize("Your paragraph goes here.")

# Initialize a WordTokenizer object
word_tokenizer = WordTokenizer()

# Tokenize a given sentence as a string
word_tokenizer.tokenize("Your sentence goes here.")

To-do List

Our to-do list includes:

  • Usage examples (Complete)
  • Cython C-API for enhanced performance (Complete, see build/tr_tokenizer.c)
  • Release platform-specific shared dynamic libraries (Complete, e.g., build/tr_tokenizer.cpython-38-x86_64-linux-gnu.so, available for Debian Linux with GCC compiler)
  • Document any limitations
  • Provide a straightforward guide for contributing

Additional Resources

Explore more about natural language processing and related topics: