from flask import Flask, request, jsonify import tensorflow as tf from transformers import AutoTokenizer, TFT5ForConditionalGeneration from transformers import MBartForConditionalGeneration, MBart50Tokenizer import os import re import spacy from nltk.corpus import wordnet as wn import random import nltk nltk.download('wordnet') nlp = spacy.load("en_core_web_sm") app = Flask(__name__) # Model uploaded configuration LOCAL_QG_MODEL_PATH = "blaxx14/t5-question-generation" """string into dictionary""" def parse_to_dict(input_string): try: question_part, answer_part = input_string.split('Answer: ') question = question_part.replace('Question: ', '').strip() answer = answer_part.strip() result_dict = { "Question": question, "Answer": answer } return result_dict except ValueError: print("Format input string tidak sesuai") return None """Find sinonim""" def get_synonyms(word): synonyms = set() for syn in wn.synsets(word): for lemma in syn.lemmas(): synonyms.add(lemma.name()) return list(synonyms) """Create distractor""" def generate_distractors(question, correct_answer): doc = nlp(question) keywords = [token.text for token in doc if token.pos_ in ['NOUN', 'PROPN']] distractors = [] for keyword in keywords: synonyms = get_synonyms(keyword) synonyms = [word for word in synonyms if word.lower() != correct_answer.lower()] distractors.extend(synonyms) distractors = random.sample(distractors, min(3, len(distractors))) return distractors """Load question generator model and tokenizer""" print("Loading model...") model = TFT5ForConditionalGeneration.from_pretrained(LOCAL_QG_MODEL_PATH, from_pt=False) tokenizer = AutoTokenizer.from_pretrained("t5-small") print("Model loaded successfully.") """Function for generate question""" def generate_question(text, max_length=4096): input_text = f"Generate question answer: {text}" input_ids = tokenizer.encode(input_text, return_tensors="tf", max_length=512, truncation=True) output = model.generate( input_ids, max_length=max_length, num_beams=10, top_k=0, top_p=0.8, temperature=1.5, do_sample=True, early_stopping=True ) output_text = tokenizer.decode(output[0], skip_special_tokens=True) return output_text """Cleaning input""" def clean_text(text): cleaned_text = text.replace("translit.", "") cleaned_text = re.sub(r'\[.*?\]', '', cleaned_text) return cleaned_text def split_text_into_sentences(paragraph): text = clean_text(paragraph) sentences = re.split(r'(?<=[.?!])\s+', text) return sentences def split_into_parts(sentences, num_parts=5): if len(sentences) <= num_parts: return sentences else: part_size = len(sentences) // num_parts parts = [sentences[i:i + part_size] for i in range(0, len(sentences), part_size)] if len(parts) > num_parts: parts[-2].extend(parts[-1]) parts = parts[:-1] return parts """Route for run generator and save the results in cloud""" @app.route('/generate-question', methods=['POST']) def api_generate_question(): try: data = request.json text = data.get('text', '') if not text: return jsonify({'error': 'Text tidak boleh kosong'}), 400 """Run cleaning input""" formatted_sentences = split_text_into_sentences(text) parts = split_into_parts(formatted_sentences) """Just for checking""" #print(parts) """Generate question""" question_list = [] for sentence in parts: combined_input = ' '.join(sentence) result = generate_question(combined_input) result_dict = parse_to_dict(result) # print(result_dict) distractors = generate_distractors(result_dict["Question"], result_dict["Answer"]) result_dict["distractor"] = distractors question_list.append(result_dict) print(question_list) return jsonify({'generated_question': question_list}) except Exception as e: return jsonify({'error': str(e)}), 500 if __name__ == '__main__': app.run(host='0.0.0.0', port=8080)