app.py

import re
import numpy as np
import pandas as pd
import nltk
import langdetect
from nltk.corpus import stopwords
from nltk.stem import WordNetLemmatizer

import gradio as gr

from sklearn.model_selection import train_test_split
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import accuracy_score, f1_score

from tensorflow.keras.preprocessing.sequence import pad_sequences
from tensorflow.keras.preprocessing.text import Tokenizer
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Embedding, LSTM, Dense

from transformers import pipeline

# -----------------------------
# 1. NLTK деректерін жүктеу
# -----------------------------
nltk.download('stopwords')
nltk.download('wordnet')

# -----------------------------
# 2. Тіл анықтау
# -----------------------------
def detect_language(text):
    try:
        lang = langdetect.detect(text)
        if lang == "ru":
            return "Russian"
        elif lang == "en":
            return "English"
        elif lang == "kk":
            return "Kazakh"
        else:
            return "Unknown"
    except:
        return "Unknown"

# -----------------------------
# 3. Текстті тазалау
# -----------------------------
stop_words_en = set(stopwords.words("english"))
lemm = WordNetLemmatizer()

def clean_text(text):
    text = text.lower()
    text = re.sub(r"http\S+", "", text)
    text = re.sub(r"[^a-z ]", "", text)
    tokens = text.split()
    tokens = [lemm.lemmatize(w) for w in tokens if w not in stop_words_en]
    return " ".join(tokens)

# -----------------------------
# 4. Demo Dataset
# -----------------------------
data = {
    "text": [
        "I love this movie!",
        "Terrible experience.",
        "It is okay.",
        "Absolutely wonderful!",
        "Worst product ever!",
        "Not bad at all.",
        "I am happy.",
        "I am angry."
    ],
    "label": [1,0,1,1,0,1,1,0]
}

df = pd.DataFrame(data)
df["clean"] = df["text"].apply(clean_text)

X = df["clean"]
y = df["label"]

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.25)

# -----------------------------
# 5. Logistic Regression
# -----------------------------
tfidf = TfidfVectorizer()
X_train_tfidf = tfidf.fit_transform(X_train)

log_reg = LogisticRegression()
log_reg.fit(X_train_tfidf, y_train)

# -----------------------------
# 6. LSTM Model
# -----------------------------
tokenizer = Tokenizer()
tokenizer.fit_on_texts(X_train)

X_train_seq = tokenizer.texts_to_sequences(X_train)
max_len = 20
X_train_pad = pad_sequences(X_train_seq, maxlen=max_len)

lstm = Sequential()
lstm.add(Embedding(input_dim=len(tokenizer.word_index)+1, output_dim=32))  # input_length алып тасталды
lstm.add(LSTM(32))
lstm.add(Dense(1, activation="sigmoid"))
lstm.compile(optimizer="adam", loss="binary_crossentropy", metrics=["accuracy"])
lstm.fit(X_train_pad, y_train, epochs=3, batch_size=4, verbose=0)

# -----------------------------
# 7. BERT Pipeline (CPU)
# -----------------------------
bert_model = pipeline(
    "sentiment-analysis",
    model="nlptown/bert-base-multilingual-uncased-sentiment",
    device=-1  # CPU режимінде
)

# -----------------------------
# 8. Prediction function
# -----------------------------
def analyze_text(text):
    lang = detect_language(text)
    cleaned = clean_text(text)

    # Logistic Regression
    vec = tfidf.transform([cleaned])
    pred_lr = log_reg.predict(vec)[0]
    label_lr = "Positive 😊" if pred_lr == 1 else "Negative 😡"

    # LSTM
    seq = tokenizer.texts_to_sequences([cleaned])
    pad = pad_sequences(seq, maxlen=max_len)
    pred_lstm = (lstm.predict(pad)[0][0] > 0.5).astype(int)
    label_lstm = "Positive 😊" if pred_lstm == 1 else "Negative 😡"

    # BERT
    res = bert_model(text)[0]["label"]
    label_bert = "Positive 😊" if res in ["4 stars","5 stars"] else "Negative 😡"

    return {
        "Detected Language": lang,
        "Logistic Regression": label_lr,
        "LSTM (Keras)": label_lstm,
        "BERT": label_bert
    }

# -----------------------------
# 9. Gradio Interface
# -----------------------------
ui = gr.Interface(
    fn=analyze_text,
    inputs=gr.Textbox(label="Enter text / Введите текст"),
    outputs=gr.JSON(label="Results / Результаты"),
    title="Multilingual Sentiment Analysis",
    description="Supports English, Russian, Kazakh. Автоматически определяет язык."
)

ui.launch()