nyansasua 0.1.0

Multi-language NLP keyword extraction (C++17 core, pybind11 bindings).

Nyansasua: Blazing-fast Multi-Language Keyword Extraction

Nyansasua (Twi) — learning / wisdom.

A self-contained, high-performance Python library for multi-language keyword extraction, backed by a low-level C++17 core.

Features

• 14 Languages Supported: English, Spanish, French, German, Italian, Portuguese, Dutch, Russian, Chinese, Japanese, Korean, Arabic, Indonesian, and Twi.
• 4 Extraction Algorithms: TF-IDF, YAKE, TextRank, and RAKE. Also supports a combined Ensemble mode.
• Lightning Fast: Highly optimized C++ backend. Benchmarks show processing of 1.5+ million characters (~200,000 words) in < 0.5 seconds.
• UTF-8 Native: Natively handles complex Unicode and mixed-script texts without breaking token boundaries.

Installation

pip install nyansasua

Quick Start

import cire

# One-liner extraction
for k in cire.extract_keywords("Machine learning is a branch of AI.", top_k=5):
    print(k.text, k.score)

# Using the Extractor class
ext = cire.Extractor(language="auto", algorithm="ensemble", top_k=10)
for k in ext.extract("Natural language processing has seen rapid growth."):
    print(k.text, k.score)

Batch Processing & TF-IDF Extraction

import cire

ext = cire.Extractor(language="auto")

# Extract from a batch of documents
results = ext.extract_many([
    "Python is widely used in data science.",
    "Climate change is a significant global challenge."
])

# Corpus-driven TF-IDF extraction
corpus = [
    "Python is used in data science.", 
    "Java is used in enterprise environments.",
    "Python is popular for AI."
]
kws = ext.extract_corpus_tfidf(
    texts=corpus, 
    target_text="Python is heavily utilized in AI and ML.", 
    top_k=3
)

Language Detection & Stopwords

import cire

# Detect the dominant script heuristically
lang = cire.detect_language("Bonjour tout le monde") 
print(lang)  # Language.French

# Add custom stopwords dynamically
cire.add_stopword("french", "tout")

License

MIT License.