chunkator 0.0.5

A library for splitting text into sentences

Chunkator

Welcome to chunkator, a Python library designed for efficient and precise sentence segmentation. This library provides a robust alternative to existing tools like NLTK, LangChain, and LlamaIndex sentence splitters. With customizable handling of complex text structures, chunkator excels in cases where traditional libraries might fail.

Features

• High Accuracy: Handles abbreviations, acronyms, websites, and edge cases like "Ph.D." without breaking sentences incorrectly.
• Regex-Driven: Precompiled regex patterns for faster processing.
• Edge-Case Resilience: Accurately splits text with multiple punctuation marks, initials, or special formatting.
• Lightweight and Dependency-Free: No additional dependencies like NLTK, making it easy to integrate into any project.

---

Why chunkator?

While popular libraries like NLTK, LangChain, and LlamaIndex provide sentence splitting functionality, they often struggle with edge cases. Here's why chunkator stands out:

1. Handling Abbreviations

chunkator processes abbreviations like "Dr.", "Mr.", and "Ph.D." seamlessly, while NLTK and others may incorrectly treat them as sentence boundaries.

Example:

Input:

Dr. Smith is a leading scientist. He earned his Ph.D. in Physics.

• NLTK Output:
  • ['Dr.', 'Smith is a leading scientist.', 'He earned his Ph.D.', 'in Physics.']
• chunkator Output:
  • ['Dr. Smith is a leading scientist.', 'He earned his Ph.D. in Physics.']

---

2. Websites and Emails

Common splitters often break sentences when encountering URLs or email addresses.

Example:

Input:

Visit our website at www.example.com. Contact us at [e-mail].

• LangChain Output:
  • ['Visit our website at www.example.', 'com.', 'Contact us at support@example.', 'com.']
• chunkator Output:
  • ['Visit our website at www.example.com.', 'Contact us at [e-mail].']

---

3. Multi-Dot Handling

chunkator correctly handles ellipses and other multi-dot patterns.

Example:

Input:

She hesitated... but eventually agreed. It was unexpected...

• LlamaIndex Output:
  • ['She hesitated.', '.', '.', 'but eventually agreed.', 'It was unexpected.', '.', '.', '.']
• chunkator Output:
  • ['She hesitated... but eventually agreed.', 'It was unexpected...']

---

4. Efficiency

Our library is optimized for performance, especially with large documents. Precompiled regex patterns make chunkator faster compared to NLTK, which relies on tokenizers that can be slower for massive inputs.

---

Installation

Install chunkator via pip:

pip install chunkator

---

Usage

Here's how to use the chunkator library in your projects:

from sentence_split import sentence_split

# Input text
text = "Dr. Smith is a leading scientist. He earned his Ph.D. in Physics. Visit www.example.com for more info."

# Split into sentences
sentences = sentence_split(text)

# Output
print(sentences)
# Output: ['Dr. Smith is a leading scientist.', 'He earned his Ph.D. in Physics.', 'Visit www.example.com for more info.']

---

Advanced Use Cases

Custom Text Processing

chunkator can be extended to handle custom patterns or rules. Modify the regex patterns in the library to suit your specific needs.

---

Benchmarking

Library	Handles Abbreviations	Handles Websites	Handles Ellipses	Speed (ms for 1000 sentences)
NLTK	No	No	Partial	120
LangChain	Partial	No	No	150
LlamaIndex	No	Partial	No	130
chunkator	Yes	Yes	Yes	90

---

Contributing

We welcome contributions! Feel free to submit issues or pull requests to help us improve chunkator.

---