app.py

"""
Hotel Review Analysis and Response System
ISOM5240 Group Project
Automatically analyzes hotel guest reviews in multiple languages, performs sentiment
analysis and aspect detection, then generates professional responses.
"""

import streamlit as st
from transformers import (
    pipeline,
    AutoModelForSequenceClassification,
    AutoTokenizer
)
import torch
import re
from langdetect import detect

# ===== CONSTANTS =====
MAX_CHARS = 1000  # Character limit for reviews

# Supported languages with their display names
SUPPORTED_LANGUAGES = {
    'en': 'English',
    'zh': 'Chinese',
    'ja': 'Japanese',
    'ko': 'Korean',
    'fr': 'French',
    'de': 'German'
}

# ===== ASPECT CONFIGURATION =====
aspect_map = {
    # Location related
    "location": ["location", "near", "close", "access", "transport", "distance", "area", "tsim sha tsui", "kowloon"],
    "view": ["view", "scenery", "vista", "panorama", "outlook", "skyline"],
    "parking": ["parking", "valet", "garage", "car park", "vehicle"],

    # Room related
    "room comfort": ["comfortable", "bed", "pillows", "mattress", "linens", "cozy", "hard", "soft"],
    "room cleanliness": ["clean", "dirty", "spotless", "stains", "hygiene", "sanitation", "dusty"],
    "room amenities": ["amenities", "minibar", "coffee", "tea", "fridge", "facilities", "tv", "kettle"],
    "bathroom": ["bathroom", "shower", "toilet", "sink", "towel", "faucet", "toiletries"],

    # Service related
    "staff service": ["staff", "friendly", "helpful", "rude", "welcoming", "employee", "manager"],
    "reception": ["reception", "check-in", "check-out", "front desk", "welcome", "registration"],
    "housekeeping": ["housekeeping", "maid", "cleaning", "towels", "service", "turndown"],
    "concierge": ["concierge", "recommendation", "advice", "tips", "guidance", "directions"],
    "room service": ["room service", "food delivery", "order", "meal", "tray"],

    # Facilities
    "dining": ["breakfast", "dinner", "restaurant", "meal", "food", "buffet", "lunch"],
    "bar": ["bar", "drinks", "cocktail", "wine", "lounge", "happy hour"],
    "pool": ["pool", "swimming", "jacuzzi", "sun lounger", "deck", "towels"],
    "spa": ["spa", "massage", "treatment", "relax", "wellness", "sauna"],
    "fitness": ["gym", "fitness", "exercise", "workout", "training", "weights"],

    # Technical
    "Wi-Fi": ["wifi", "internet", "connection", "online", "network", "speed"],
    "AC": ["air conditioning", "AC", "temperature", "heating", "cooling", "ventilation"],
    "elevator": ["elevator", "lift", "escalator", "vertical transport", "wait"],

    # Value
    "pricing": ["price", "expensive", "cheap", "value", "rate", "cost", "worth"],
    "extra charges": ["charge", "fee", "bill", "surcharge", "additional", "hidden"]
}

aspect_responses = {
    "location": "We're delighted you enjoyed our prime location in the heart of Tsim Sha Tsui with convenient access to major attractions.",
    "view": "It's wonderful to hear you appreciated the stunning views of Victoria Harbour from your room.",
    "room comfort": "Our team is thrilled you found your room comfortable and well-appointed for your needs.",
    "room cleanliness": "Your commendation of our cleanliness standards means a great deal to our housekeeping team who work diligently to maintain our high standards.",
    "staff service": "Your kind words about our team have been shared with them and are greatly appreciated.",
    "reception": "We're pleased our front desk team made your arrival and departure experience seamless and welcoming.",
    "spa": "Our spa practitioners will be delighted you enjoyed their treatments and the relaxing ambiance of our wellness center.",
    "pool": "We're glad you had a refreshing time at our rooftop pool with its panoramic city views.",
    "dining": "Thank you for appreciating our culinary offerings - we've shared your compliments with our executive chef and culinary team.",
    "concierge": "We're happy our concierge could enhance your stay with their local knowledge and personalized recommendations.",
    "fitness": "It's great to hear you made use of our 24-hour fitness center with its modern equipment.",
    "room service": "We're pleased our in-room dining met your expectations for both quality and timely service.",
    "parking": "We're glad our valet parking service provided convenience during your stay with us.",
    "bathroom": "We appreciate your feedback about our bathroom amenities and the cleanliness of your facilities.",
    "bar": "Thank you for your comments about our bar service and the selection of beverages available in our lounge.",
    "housekeeping": "Your positive feedback about our housekeeping service has been shared with the entire team.",
    "Wi-Fi": "We're pleased our high-speed internet service met your connectivity needs throughout the property.",
    "elevator": "We're glad our elevator service provided convenient access to all areas of the hotel during your stay."
}

improvement_actions = {
    "AC": "completed a comprehensive inspection and maintenance of all air conditioning units",
    "housekeeping": "conducted additional training for our housekeeping team and adjusted cleaning schedules",
    "bathroom": "performed deep cleaning and maintenance on all bathroom facilities",
    "parking": "implemented enhanced key management protocols with our valet service team",
    "dining": "reviewed our menu pricing and quality standards with the culinary leadership team",
    "reception": "provided additional customer service training to our front desk associates",
    "elevator": "completed full servicing and testing of all elevator systems",
    "room amenities": "begun upgrading in-room amenities based on recent guest feedback",
    "noise": "initiated soundproofing improvements in identified high-traffic areas",
    "pricing": "commenced a comprehensive review of our pricing structure and value proposition",
    "Wi-Fi": "begun upgrading our network infrastructure to enhance connectivity",
    "bar": "reviewed our beverage service procedures and inventory management",
    "staff service": "implemented additional staff training programs focusing on guest interactions",
    "room service": "optimized our food delivery processes to improve efficiency",
    "fitness": "scheduled upgrades to our gym equipment based on guest preferences"
}

# ===== MODEL CONFIGURATION =====
TRANSLATION_MODELS = {
    # Translations to English
    'zh-en': 'Helsinki-NLP/opus-mt-zh-en',
    'ja-en': 'Helsinki-NLP/opus-mt-ja-en',
    'ko-en': 'Helsinki-NLP/opus-mt-ko-en',
    'fr-en': 'Helsinki-NLP/opus-mt-fr-en',
    'de-en': 'Helsinki-NLP/opus-mt-de-en',
    
    # Translations from English
    'en-zh': 'Helsinki-NLP/opus-mt-en-zh',
    'en-ja': 'Helsinki-NLP/opus-mt-en-ja',
    'en-ko': 'Helsinki-NLP/opus-mt-en-ko',
    'en-fr': 'Helsinki-NLP/opus-mt-en-fr',
    'en-de': 'Helsinki-NLP/opus-mt-en-de'
}

# ===== MODEL LOADING =====
@st.cache_resource
def load_sentiment_model():
    model = AutoModelForSequenceClassification.from_pretrained("smtsead/fine_tuned_bertweet_hotel")
    tokenizer = AutoTokenizer.from_pretrained('finiteautomata/bertweet-base-sentiment-analysis')
    return model, tokenizer

@st.cache_resource
def load_aspect_classifier():
    return pipeline("zero-shot-classification", model="MoritzLaurer/deberta-v3-base-zeroshot-v1.1-all-33")

@st.cache_resource
def load_translation_model(src_lang, target_lang='en'):
    model_key = f"{src_lang}-{target_lang}"
    if model_key not in TRANSLATION_MODELS:
        raise ValueError(f"Unsupported translation: {src_lang}→{target_lang}")
    return pipeline("translation", model=TRANSLATION_MODELS[model_key])

# ===== CORE FUNCTIONS =====
def detect_language(text):
    try:
        lang = detect(text)
        return 'zh' if lang in ['zh', 'yue'] else lang if lang in SUPPORTED_LANGUAGES else 'en'
    except:
        return 'en'

def translate_text(text, src_lang, target_lang='en'):
    try:
        if src_lang == target_lang:
            return {'translation': text, 'source_lang': src_lang}
        translator = load_translation_model(src_lang, target_lang)
        result = translator(text)[0]['translation_text']
        return {'translation': result, 'source_lang': src_lang}
    except Exception as e:
        return {'error': str(e)}

def analyze_sentiment(text, model, tokenizer):
    inputs = tokenizer(text, padding=True, truncation=True, max_length=512, return_tensors='pt')
    with torch.no_grad():
        outputs = model(**inputs)
        probs = torch.nn.functional.softmax(outputs.logits, dim=-1)
        predicted_label = torch.argmax(probs).item()
        confidence = torch.max(probs).item()
    return {
        'label': predicted_label,
        'confidence': f"{confidence:.0%}",
        'sentiment': 'POSITIVE' if predicted_label else 'NEGATIVE'
    }

def detect_aspects(text, aspect_classifier):
    relevant_aspects = []
    text_lower = text.lower()
    for aspect, keywords in aspect_map.items():
        if any(re.search(rf'\b{kw}\b', text_lower) for kw in keywords):
            relevant_aspects.append(aspect)
    
    if relevant_aspects:
        result = aspect_classifier(
            text,
            candidate_labels=relevant_aspects,
            multi_label=True,
            hypothesis_template="This review discusses the hotel's {}."
        )
        return [(aspect, f"{score:.0%}") for aspect, score in zip(result['labels'], result['scores']) if score > 0.6]
    return []

def generate_response(sentiment, aspects, original_text):
    # Personalization - only extract guest name
    guest_name = ""
    name_match = re.search(r"(Mr\.|Ms\.|Mrs\.)\s(\w+)", original_text, re.IGNORECASE)
    
    if name_match:
        guest_name = f" {name_match.group(2)}"

    if sentiment['label'] == 1:
        response = f"""Dear{guest_name if guest_name else ' Valued Guest'},
Thank you for choosing our hotel and for sharing your kind feedback with us."""
        
        # Add relevant aspect responses
        added_aspects = set()
        for aspect, _ in sorted(aspects, key=lambda x: float(x[1][:-1]), reverse=True):
            if aspect in aspect_responses:
                response_text = aspect_responses[aspect]
                response += "\n\n" + response_text
                added_aspects.add(aspect)
                if len(added_aspects) >= 3:
                    break
        
        response += "\n\nWe look forward to welcoming you back for another memorable stay."
    else:
        response = f"""Dear{guest_name if guest_name else ' Guest'},
Thank you for taking the time to share your feedback with us. We sincerely regret that your experience did not meet your expectations."""
        
        # Add improvement actions
        added_improvements = set()
        improvement_text = ""
        for aspect, _ in sorted(aspects, key=lambda x: float(x[1][:-1]), reverse=True):
            if aspect in improvement_actions:
                improvement_text += f"\n- Regarding the {aspect}, we have {improvement_actions[aspect]}"
                added_improvements.add(aspect)
                if len(added_improvements) >= 2:
                    break
        
        if improvement_text:
            response += "\n\nTo address your concerns:" + improvement_text
        
        response += "\n\nYour feedback is invaluable to us as we strive to improve our services."
    
    # Common closing
    response += """
    
Should you require any further assistance, please don't hesitate to contact our Guest Relations team.
Sincerely yours,
Guest Relations Team
The Mira Hong Kong
+852 1234 5678 | [email protected]"""
    
    return response

# ===== STREAMLIT UI =====
def main():
    st.set_page_config(
        page_title="Hotel Review Analysis and Response System",
        page_icon="🏨",
        layout="centered"
    )
    
    st.markdown("""
    <style>
        .header { color: #003366; font-size: 28px; font-weight: bold; margin-bottom: 10px; }
        .subheader { color: #666666; font-size: 16px; margin-bottom: 30px; }
        .char-counter { font-size: 12px; color: #666; text-align: right; margin-top: -15px; }
        .char-counter.warning { color: #ff6b6b; }
        .result-box { border-left: 4px solid #003366; padding: 15px; background-color: #f9f9f9; margin: 20px 0; }
        .aspect-badge { background-color: #e6f2ff; padding: 2px 8px; border-radius: 4px; display: inline-block; margin: 2px; }
        .response-box { white-space: pre-wrap; font-family: monospace; }
        .english-response { color: #555555; font-size: 14px; }
    </style>
    """, unsafe_allow_html=True)
    
    st.markdown('<div class="header">Hotel Review Analysis and Response System</div>', unsafe_allow_html=True)
    st.markdown('<div class="subheader">The Mira Hong Kong</div>', unsafe_allow_html=True)
    
    review = st.text_area("**Paste Guest Review:**", 
                         height=200,
                         max_chars=MAX_CHARS,
                         placeholder=f"Enter review (max {MAX_CHARS} characters)...",
                         key="review_input")
    
    char_count = len(st.session_state.review_input) if 'review_input' in st.session_state else 0
    st.markdown(f'<div class="char-counter{" warning" if char_count > MAX_CHARS else ""}">{char_count}/{MAX_CHARS} characters</div>', 
                unsafe_allow_html=True)
    
    if st.button("Analyze & Generate Response", type="primary"):
        if not review.strip():
            st.error("Please enter a review")
            return
        
        if char_count > MAX_CHARS:
            st.warning(f"Review truncated to {MAX_CHARS} characters")
            review = review[:MAX_CHARS]
        
        with st.spinner("Analyzing feedback..."):
            try:
                # Auto-detect language
                review_lang = detect_language(review)
                st.info(f"Detected language: {SUPPORTED_LANGUAGES.get(review_lang, 'English')}")
                
                # Translate if not English
                if review_lang != 'en':
                    translation = translate_text(review, review_lang, 'en')
                    if 'error' in translation:
                        st.error(f"Translation error: {translation['error']}")
                        return
                    analysis_text = translation['translation']
                    
                    with st.expander("View Translation"):
                        st.write("**Original Review:**")
                        st.write(review)
                        st.write("**English Translation:**")
                        st.write(translation['translation'])
                else:
                    analysis_text = review
                
                # Analyze text
                sentiment_model, tokenizer = load_sentiment_model()
                aspect_classifier = load_aspect_classifier()
                
                sentiment = analyze_sentiment(analysis_text, sentiment_model, tokenizer)
                aspects = detect_aspects(analysis_text, aspect_classifier)
                response = generate_response(sentiment, aspects, review)
                
                # Translate response back if needed
                if review_lang != 'en':
                    translation_back = translate_text(response, 'en', review_lang)
                    final_response = translation_back['translation'] if 'error' not in translation_back else response
                else:
                    final_response = response
                
                # Display results
                st.divider()
                
                col1, col2 = st.columns(2)
                with col1:
                    st.markdown("### Sentiment Analysis")
                    st.markdown(f"{'✅' if sentiment['label'] == 1 else '⚠️'} **{sentiment['sentiment']}**")
                    st.caption(f"Confidence: {sentiment['confidence']}")
                
                with col2:
                    st.markdown("### Key Aspects")
                    if aspects:
                        for aspect, score in sorted(aspects, key=lambda x: float(x[1][:-1]), reverse=True):
                            st.markdown(f'<div class="aspect-badge">{aspect} ({score})</div>', unsafe_allow_html=True)
                    else:
                        st.markdown("_No specific aspects detected_")
                
                st.divider()
                st.markdown("### Draft Response")
                
                # Show English response if original language wasn't English
                if review_lang != 'en':
                    st.markdown('<div class="english-response">English version:</div>', unsafe_allow_html=True)
                    st.markdown(f'<div class="result-box"><div class="response-box">{response}</div></div>', 
                               unsafe_allow_html=True)
                    st.markdown('<div class="english-response">Translated version:</div>', unsafe_allow_html=True)
                
                # Show final response (translated if needed)
                st.markdown(f'<div class="result-box"><div class="response-box">{final_response}</div></div>', 
                           unsafe_allow_html=True)
                
            except Exception as e:
                st.error(f"An error occurred: {str(e)}")

if __name__ == "__main__":
    main()