keyphrase_extraction.py

from rake_nltk import Rake
import nltk
import re

# Download NLTK data
nltk.download('punkt')
nltk.download('stopwords')

# Define a list of obligation words
obligation_words = [
    # English words
    "must", "will", "use", "may", "provides", 'is obliged to', 
    'has to', 'needs to', 'is required to', "shall", 
    "should", "ought to", "required", "obligated", "duty", 
    "agrees to", "has a duty to", "is expected to", "commits to",
    
    # German words 
    "muss", "wird", "nutzen", "darf", "stellt bereit", 
    "ist verpflichtet", "ist erforderlich", "soll", 
    "sollte", "erforderlich", "verpflichtet", "Pflicht", 
    "stimmt zu", "hat die Pflicht", "wird erwartet", "verpflichtet sich"
]


def extract_sentences_with_obligations(text):
    # Initialize Rake with stopwords set to None (to keep all words)
    rake = Rake()

    # Split the text into sentences
    sentences = re.split(r'(?<=[.!?])\s+', text)

    # Initialize a list to store sentences with obligation words
    obligation_sentences = []

    # Iterate through the sentences
    for sentence in sentences:
        # Extract keyphrases from the sentence
        rake.extract_keywords_from_text(sentence)
        
        # Get the ranked keyphrases
        ranked_keyphrases = rake.get_ranked_phrases()

        # Check if any of the ranked keyphrases contain obligation words
        if any(any(word in kp.lower() for word in obligation_words) for kp in ranked_keyphrases):
            obligation_sentences.append(sentence)

    # Join the sentences into a single string
    return ' '.join(obligation_sentences)