This projects aims to build a library for all the NLP processes for Nepali Language.

Getting the module

!pip install git+https://github.com/sushil79g/Nepali_nlp.git

Loading Embedding

from Nepali_nlp import Embeddings
word_vec = Embeddings().load_large_vector()
#word_vec = Embeddings().load_vector() #For small Embedding
#from fasttext_embedding import Fasttext
#word_vec = Fasttext().load()

For Nepali Synonym

from Nepali_nlp import Synonym
Synonym().raw_synonym(word = 'माया',word_vec=word_vec) #method: 1
#output -> स्नेह','प्रेम','आदर','मायाँ','दया','मायालु','श्रद्धा','आत्मियता','स्पर्श','तिमी
Synonym().filter_synonym(word = 'साथी',word_vec=word_vec) #method: 2
#output -> 'भाइहरू','सहपाठी','प्रेमी','दाइ','प्रेमि','बहिनी'

Word-spell corrector

from Nepali_nlp import Corrector
Corrector().corrector(word='सुशल') #In a very raw stage for now.
#output-> ['सुशील', 'सुशील']
Corrector().spell_correct("कस्त भको हेरौ है")
#output-> "कस्तो भयो हेर है"

Nepali text summerizer

from Nepali_nlp import Summerize
Summerize().show_summary(word_vec,text, length_sentence_predict=5)

Nepali unicode to Devnagiri Font

from Nepali_nlp import Unicode
text = 'ma ghara jaanchhu'
Unicode().unicode_word(text) #output-> 'म घर जान्छु'

Preeti-font character to Devnagiri Font

from Nepali_nlp import preeti
unicode_word = 'g]kfnL'
print(preeti(unicode_word)) #output-> नेपाली

OCR(optical character reader)

from Nepali_nlp import OCR
text = OCR(image_location)

Nepali Tokenizer

from Nepali_nlp import Tokenizer
Tokenizer().sentence_tokenize(text) #To tokenize sentence
Tokenizer().word_tokenize(text) #To tokenize word
Tokenizer().character_tokenize(text) #To tokenize character

Nepali sentence similarity

from Nepali_nlp import  Avg_vector_similar
sentences = ["कुपोषणकै कारण शारीरिक र मानसिक रुपमा कमजोर मात्र होइन, अकालमै ज्यान पनि गुमाउनुको परेको समाचार बग्रेल्ती सुन्न सकिन्छ","कर्णाली प्रदेश सामाजिक विकास मन्त्रालयले उपलब्ध गराएको तथ्यांकले कर्णालीमा प्रत्येक वर्ष जन्मिएका ५ वर्षमुनीका बालबालिका १ हजार जनामध्ये ५८ जनाले ज्यान गुमाउँदै आएको देखाएको छ"]
Avg_vector_similar().pair_similarity(word_vec, sentences) #output-> 0.6817289590835571

Nepali new-portal Scrapper (onlinekhabar and ekantipur for now)

from Nepali_nlp import extract_news
news_link = 'https://www.onlinekhabar.com/2019/12/821094'
title, news = extract_news(news_link) #onlinekhabar and ekantipur is supported at the moment.

Show latest news summary

from Nepali_nlp import UpdateNews
title, links, summerized_news = UpdateNews().show_latest(word_vec=word_vec,portal='onlinekhabar',number_of_news=5) #ekantipur portal is also supported

TODOs:

• Nepali Embeddings
• Tokenizers (sentence, word, character)
• Stop Words
• Nepali Words Collection
• Nepali Word synonym
• Roman Nepali to Nepali
• Nepali OCR
• Summerization
• Pos_tag
• Sentence similarity score
• Spell correction
• Named Entity Recognition (Currently)
• Translation(Nepali<->English)