aamraz 0.0.9

This project is a collection of Natural Language Processing tools for Kurdish Language.

Aamraz - Kurdish NLP collection

Overview

Aamraz which is written "ئامراز" in kurdish script means "instrument". This project is a collection of Natural Language Processing tools for Kurdish Language. Despite being spoken by millions, Kurdish remains an under-resourced language in Natural Language Processing (NLP). Recognizing the rich cultural heritage and historical significance of the Kurdish people, we—regardless of ethnicity—are committed to advancing tools and pre-trained models that empower the Kurdish language in modern research and technology. Our work aims to foster further development and provide a foundation for future research and applications in NLP.

Base Features

• Normalization
• Tokenization
• Stemming
• Word Embedding: Creates vector representations of words.
• Sentences Embedding: Creates vector representations of sentences.

Tools

Installation

pip install aamraz

PretrainedModels

some useful pre-trained Models:

Model	Version	Description	Size
FastText WordEmbedding	2	Model trained using FastText method on our own Corpus. This is bot the fasttext & skip-gram model itself (fasttext model.	~ 2.3 GB
FastText WordEmbedding - Lite	1	Model trained using FastText method on our own Corpus. This is bot the fasttext & skip-gram model itself (fasttext model.	~ 800 MB
Word2vec Model	1	Including needed .bin and .npy files. Find other vector sizes Here	~ 92 MB

Usage

import aamraz

# Normalization
normalizer= aamraz.Normalizer()
sample_sentence="قڵبە‌کە‌م‌ بە‌  کوردی‌  قسە‌ دە‌کات‌."
normalized_sentence=normalizer.normalize(sample_sentence)
print(normalized_sentence)

# Tokenization
tokenizer = aamraz.WordTokenizer()
sample_sentence="زوانی له دربره"
tokens = tokenizer.tokenize(sample_sentence)
print(tokens)

# Embedding by fasttext
model_path = 'kurdish_fasttext_skipgram_dim300_v1.bin'
embedding_model = aamraz.EmbeddingModel(model_path, dim=50)

sample_word="ئامراز"
sample_sentence="زوانی له دربره"

word_vector = embedding_model.word_embedding(sample_word)
sentence_vector = embedding_model.sentence_embedding(sample_sentence)

print(word_vector)
print(sentence_vector)

# Embedding by word2vec
model_path = 'kurdish_word2vec_model_dim100_v1.bin'
embedding_model = aamraz.EmbeddingModel(model_path, type='word2vec')

sample_word="ئامراز"
sample_sentence="زوانی له دربره"

word_vector = embedding_model.word_embedding(sample_word)
sentence_vector = embedding_model.sentence_embedding(sample_sentence)

print(word_vector)
print(sentence_vector)

# Stemming
stemmer=aamraz.Stemmer(method='simple')
stemmed=stemmer.stem("کتێبەکانمان")
print(stemmed)