app.py

import gradio as gr
import numpy as np
from tensorflow.keras.models import load_model
from tensorflow.keras.preprocessing.sequence import pad_sequences
from sklearn.preprocessing import StandardScaler
import json
import re
from konlpy.tag import Okt
from tensorflow.keras.preprocessing.text import tokenizer_from_json
import pickle
import os
import logging

# 로그 설정
logging.basicConfig(filename='app.log', level=logging.DEBUG, format='%(asctime)s:%(levelname)s:%(message)s')

# 환경 변수 설정
os.environ['JAVA_HOME'] = '/usr/lib/jvm/java-11-openjdk-amd64'
os.environ['PATH'] = os.environ['JAVA_HOME'] + '/bin:' + os.environ['PATH']

# 모델 및 토크나이저 파일 로드
try:
    model = load_model('deep_learning_model(okt_drop).h5', compile=False)
    logging.info("Model loaded successfully.")

    with open('tokenizer(okt_drop).json', 'r', encoding='utf-8') as f:
        tokenizer_data = f.read()
    tokenizer = tokenizer_from_json(tokenizer_data)
    logging.info("Tokenizer loaded successfully.")

    with open('scaler.pkl', 'rb') as f:
        scaler = pickle.load(f)
    logging.info("Scaler loaded successfully.")

except Exception as e:
    logging.error("Error loading model, tokenizer, or scaler: %s", str(e))
    raise e

def calculate_sentence_stats(paragraph):
    paragraph = re.sub(r'\.{2,}', '.', paragraph)
    sentences = re.split(r'[.!?]', paragraph)
    sentence_lengths = [len(s.strip()) for s in sentences if s.strip()]
    sentence_count = len(sentence_lengths)
    average_length = sum(sentence_lengths) / len(sentence_lengths) if sentence_lengths else 0
    return sentence_count, average_length

def process_text(text):
    try:
        okt = Okt()
        texts = ' '.join(okt.nouns(text))
        sequences = tokenizer.texts_to_sequences([texts])
        max_len = 301
        X = pad_sequences(sequences, maxlen=max_len)
        return X
    except Exception as e:
        logging.error("Error processing text: %s", str(e))
        raise e

def predict_text(text, grade):
    try:
        X = process_text(text)
        sentence_count, sentence_average = calculate_sentence_stats(text)
        length = len(text)
        emoticon = 0
        numeric_features = np.array([[int(grade), length, emoticon, sentence_count, sentence_average]])
        numeric_features = scaler.transform(numeric_features)
        prediction = model.predict([X, numeric_features])
        predicted_label = '인공지능이 생성한 독서감상문입니다.' if prediction[0][0] > 0.5 else '사람이 작성한 독서감상문입니다.'
        return predicted_label
    except Exception as e:
        logging.error("Error predicting text: %s", str(e))
        raise e

iface = gr.Interface(
    fn=predict_text,
    inputs=[gr.Textbox(lines=10, placeholder="Enter Text Here..."), gr.Textbox(label="Grade")],
    outputs="text",
    title="독서감상문 분석기",
    description="이 독서감상문이 학생에 의해 작성되었는지, 인공지능에 의해 생성되었는지 분석합니다."
)
iface.launch(debug=True)