Spaces:
Sleeping
Sleeping
File size: 136,443 Bytes
09a64f4 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 |
# -*- coding: utf-8 -*-
"""Copy of Mona_Raghad_Lujin_projcect.ipynb
Automatically generated by Colab.
Original file is located at
https://colab.research.google.com/drive/1qn3CfGuoWA8ghVrJ2l8KifSOj7A2RI6J
# Smart Travel: Discover Your Perfect Hotel Using AI ✈
**Objective:**
The Project Search Agent is designed to revolutionize the way users find and interact with information by leveraging advanced artificial intelligence (AI) and natural language processing (NLP) technologies. The search agent aims to provide intelligent, personalized, and efficient search experiences across various domains, such as hotels, products, or content, by understanding user intent and delivering relevant results.
**Architecture and Workflow:**
**Data Ingestion and Preparation:**
Datasets: Utilize datasets containing structured and unstructured data related to user queries, items, or subjects of interest (e.g., hotel reviews, product descriptions).
Data Cleaning: Implement data preprocessing using pandas to handle missing values, remove noise, and standardize text data for further analysis.
**Natural Language Processing (NLP):**
Text Preprocessing: Use libraries like nltk or spaCy to tokenize, lemmatize, and remove stopwords from the text data, ensuring it is clean and consistent for analysis.
Sentiment Analysis: Apply sentiment analysis models to understand user sentiments and extract meaningful insights that guide the search recommendations.
**transformers Models:**
Semantic Search: Employ models from the transformers library, such as BERT or Sentence Transformers, to convert text into embeddings for semantic analysis.
Similarity Calculation: Use sklearn.metrics.pairwise to compute cosine similarity between query embeddings and item embeddings, identifying the most relevant matches.
**Recommendation System:**
Content-Based Filtering: Recommend items based on the semantic similarity of user queries to item attributes or reviews.
Personalized Suggestions: Incorporate user preferences and feedback to refine recommendations, ensuring a personalized experience.
**Data Visualization:**
Visual Insights: Use matplotlib and seaborn to create visual representations of search results, sentiment trends, and item ratings.
Comparative Analysis: Provide visual comparisons to help users understand the context and relevance of recommended items.
**Interactive User Interface:**
Gradio Integration: Develop an interactive web-based interface using Gradio, allowing users to input queries and view results in real-time.
User Interaction: Enable seamless interaction and instant feedback through an intuitive, user-friendly design that facilitates exploration and discovery.
# Installing Required Libraries
"""
!pip install datasets
!pip install keyphrase-vectorizers
!pip install transformers
!pip install sentence-transformers
!pip install gradio
"""# Data Preparation and Cleaning"""
import pandas as pd
df = pd.DataFrame(ds["train"])
df.info()
missing_values_per_column = df.isnull().sum()
print("Missing values per column:")
print(missing_values_per_column)
import pandas as pd
# Assuming df is your DataFrame
# Display initial missing value counts
print("Initial missing values per column:")
print(df.isnull().sum())
# Fill missing values in 'review_text' with an empty string and 'rate' with a specific value
df['review_text'].fillna('', inplace=True)
df['rate'].fillna(df['rate'].mean(), inplace=True) # Filling with mean for example
# Fill missing values in 'hotel_description' with a placeholder
df['hotel_description'].fillna('Description not available', inplace=True)
# Verify the cleaning
print("\nMissing values after cleaning:")
print(df.isnull().sum())
# Display the cleaned DataFrame (first few rows)
print("\nCleaned DataFrame sample:")
print(df.head())
df.hotel_name.value_counts()
import re
import nltk
from nltk.corpus import stopwords
from nltk.tokenize import word_tokenize
from nltk.stem import WordNetLemmatizer
import pandas as pd
# Download necessary NLTK resources
nltk.download('punkt')
nltk.download('stopwords')
nltk.download('wordnet')
# Load your DataFrame (assuming it's named df)
# Ensure that df is defined elsewhere in your code with the appropriate data
# Normalize country names
df['country'] = df['country'].replace({'Türkiye': 'Turkiye'})
# Check the value counts to verify the replacement
print(df['country'].value_counts())
def clean_text(text):
"""
Cleans text data by removing punctuation, converting to lowercase, removing stopwords, and lemmatizing.
Args:
text: The text string to be cleaned.
Returns:
The cleaned text string.
"""
# Handle None or empty values
if pd.isnull(text):
return ""
# Lowercase the text
text = text.lower()
# Remove special characters and numbers
text = re.sub(r'[^a-zA-Z\s]', '', text)
# Tokenize the text
tokens = word_tokenize(text)
# Remove stop words
stop_words = set(stopwords.words('english'))
tokens = [word for word in tokens if word not in stop_words]
# Lemmatize the tokens
lemmatizer = WordNetLemmatizer()
tokens = [lemmatizer.lemmatize(word) for word in tokens]
# Join the tokens back into a string
cleaned_text = ' '.join(tokens)
return cleaned_text
# Apply text cleaning to the 'review_text' column
df['cleaned_reviews'] = df['review_text'].apply(clean_text)
print(df.columns)
print(df.isnull().sum())
"""# Sentiment Analysis and Visualization"""
from datasets import load_dataset
from nltk.sentiment import SentimentIntensityAnalyzer
import nltk
import matplotlib.pyplot as plt
import seaborn as sns
# Download necessary NLTK resources
nltk.download('vader_lexicon')
# Initialize SentimentIntensityAnalyzer
sia = SentimentIntensityAnalyzer()
def classify_sentiment(review):
"""
Classifies the sentiment of a review as positive, negative, or neutral.
Args:
review: The text of the review.
Returns:
A string indicating the sentiment category: 'positive', 'negative', or 'neutral'.
"""
# Handle None values
if review is None:
return 'neutral' # Or any default value you prefer
# Get sentiment scores
sentiment = sia.polarity_scores(review)
# Determine sentiment category
if sentiment['compound'] > 0.05:
return 'positive'
elif sentiment['compound'] < -0.05:
return 'negative'
else:
return 'neutral'
# Apply sentiment analysis to classify each review using the correct column name 'cleaned_reviews'
df['Sentiment'] = df['cleaned_reviews'].apply(classify_sentiment)
# Group reviews by sentiment
positive_reviews = df[df['Sentiment'] == 'positive']
neutral_reviews = df[df['Sentiment'] == 'neutral']
negative_reviews = df[df['Sentiment'] == 'negative']
# Display results in a structured table
def display_table(df, sentiment, title):
print(f"\n{title} Reviews:")
display_df = df[['hotel_name', 'cleaned_reviews', 'Sentiment']].head(5)
print(display_df.to_markdown(index=False))
# Display structured tables for each sentiment category
display_table(positive_reviews, 'positive', "Positive")
display_table(neutral_reviews, 'neutral', "Neutral")
display_table(negative_reviews, 'negative', "Negative")
"""# Analyzing and Visualizing Top-Rated Hotels in a Selected Locality"""
# Check for unique localities
unique_localities = df['locality'].unique()
print(f"Number of unique localities: {len(unique_localities)}")
print("Sample localities:", unique_localities)
# Prompt user to choose a locality
print("\nAvailable localities:")
for index, locality in enumerate(unique_localities):
print(f"{index + 1}. {locality}")
selected_index = int(input("\nSelect a locality by number: ")) - 1
selected_locality = unique_localities[selected_index]
print(f"\nYou selected: {selected_locality}")
# Filter DataFrame by the selected locality
df_selected_locality = df[df['locality'] == selected_locality]
# Analyze and plot the selected locality
positive_review_counts_locality = df_selected_locality[df_selected_locality['Sentiment'] == 'positive'].groupby('hotel_name').size().reset_index(name='Positive Review Count')
top_hotels_locality = positive_review_counts_locality.sort_values(by='Positive Review Count', ascending=False).head(15)
plt.figure(figsize=(10, 6))
ax = sns.barplot(data=top_hotels_locality, y='hotel_name', x='Positive Review Count', palette='viridis')
plt.title(f'Top Hotels in {selected_locality} with the Most Positive Reviews')
plt.xlabel('Number of Positive Reviews')
plt.ylabel('Hotel Name')
plt.xlim(0, top_hotels_locality['Positive Review Count'].max() * 1.1) # Adjust x-axis limit
plt.show()
"""# Hotel Review Summarization and Rating Visualization"""
from transformers import pipeline, AutoTokenizer, AutoModelForSeq2SeqLM
# Load the tokenizer and model explicitly
tokenizer = AutoTokenizer.from_pretrained("mrm8488/bert-mini2bert-mini-finetuned-cnn_daily_mail-summarization")
model = AutoModelForSeq2SeqLM.from_pretrained("mrm8488/bert-mini2bert-mini-finetuned-cnn_daily_mail-summarization")
# Initialize the summarization pipeline with the tokenizer and model
pipe = pipeline("summarization", model=model, tokenizer=tokenizer)
# Group reviews by hotel
hotel_reviews = df.groupby('hotel_name')['cleaned_reviews'].apply(list).to_dict()
# Generate summaries for each hotel
hotel_summaries_detailed = {}
for hotel, reviews in hotel_reviews.items():
all_reviews_text = ' '.join(reviews)
# Tokenize the input text and truncate to the model's maximum length
inputs = tokenizer(all_reviews_text, truncation=True, max_length=512, return_tensors="pt") # Adjust max_length if needed
# Summarize using the model
summary_ids = model.generate(**inputs, max_length=100) # Adjust max summary length as needed
summary = tokenizer.decode(summary_ids[0], skip_special_tokens=True)
hotel_summaries_detailed[hotel] = {
"summary": summary
}
# Print detailed summaries
for hotel, data in hotel_summaries_detailed.items():
print(f"Hotel: {hotel}")
print(f"Summary: {data['summary']}\n")
df.rating_value.value_counts()
import pandas as pd
import matplotlib.pyplot as plt
# Sample DataFrame setup with rating counts
rating_data = {
'rating_value': [4.5, 5.0, 4.0, 3.5, 3.0],
'count': [2677, 1520, 1520, 200, 80]
}
df_ratings = pd.DataFrame(rating_data)
# Calculate total count and percentage for each rating
df_ratings['percentage'] = (df_ratings['count'] / df_ratings['count'].sum()) * 100
# Plotting the pie chart
plt.figure(figsize=(8, 8))
plt.pie(df_ratings['percentage'], labels=df_ratings['rating_value'], autopct='%1.1f%%', startangle=140, colors=plt.cm.Paired.colors)
plt.title('Distribution of Hotel Ratings', fontsize=16, fontweight='bold')
plt.axis('equal') # Equal aspect ratio ensures that pie is drawn as a circle.
plt.show()
import pandas as pd
# Create a list to store the data
summary_data = []
# Iterate through the hotel summaries and extract data
for hotel, summary_info in hotel_summaries_detailed.items():
summary_data.append({
'hotel_name': hotel,
'summary': summary_info['summary']
})
# Create a DataFrame from the extracted data
summary_df = pd.DataFrame(summary_data)
# Calculate average rating for each hotel
average_ratings = df.groupby('hotel_name')['rating_value'].mean()
# Add average rating to summary_df
summary_df = summary_df.merge(average_ratings, on='hotel_name', how='left')
summary_df.rename(columns={'rating_value': 'average_rating'}, inplace=True)
# Merge hotel_description and city from original DataFrame
summary_df = summary_df.merge(df[['hotel_name', 'hotel_description', 'locality']].drop_duplicates('hotel_name'), on='hotel_name', how='left')
# Merge hotel_image from original DataFrame
# Get unique hotel images
hotel_images = df[['hotel_name', 'hotel_image']].drop_duplicates('hotel_name')
# Merge hotel images into summary_df
summary_df = summary_df.merge(hotel_images, on='hotel_name', how='left')
summary_df
# Plot locality vs average_rating using a box plot
figsize = (8, 1.2 * len(summary_df['locality'].unique()))
plt.figure(figsize=figsize)
sns.boxplot(data=summary_df, x='average_rating', y='locality', palette='Dark2')
sns.despine(top=True, right=True, bottom=True, left=True)
# Add title and labels
plt.title('Distribution of Average Hotel Ratings by Locality')
"""By using insights from the box plot, businesses can better understand customer preferences and satisfaction levels. This helps them improve service quality and meet customer expectations, leading to happier customers and stronger customer relationships. Ultimately, this benefits both the business and its customers, creating a positive outcome for everyone involved.
# Semantic Search and Similarity
"""
from sentence_transformers import SentenceTransformer, util
embedding = SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2')
#sample= summary_df[:20]
summary_df["embedding"] = summary_df["summary"].apply(lambda x: embedding.encode(x))
from sklearn.metrics.pairwise import cosine_similarity
from PIL import Image
from IPython.display import display
import requests
def search(query):
query_embedding = embedding.encode(query)
summary_df["similarity"] = summary_df["embedding"].apply(lambda x: cosine_similarity([query_embedding], [x])[0][0])
results = summary_df.sort_values("similarity", ascending=False).head(3)# we can chage top
print("Top 3 most similar hotels:")
for index, row in results.iterrows():
print(f"Hotel Name: {row['hotel_name']}")
print(f"hotel description: {row['hotel_description']}")
print(f"summary: {row['summary']}")
print(f"locality: {row['locality']}")
print(f"Similarity Score: {row['similarity']}\n")
# Open the image using Image.open() and display it
image = Image.open(requests.get(row['hotel_image'], stream=True).raw) # Assuming 'hotel_image' contains a URL
display(image.resize((200, 200))) # Resize and display the image
print("\n")
# Example usage
query = 'a hotel in london and great food nearby but not too expensive'
q_cleaned = clean_text(query)
results = search(q_cleaned)
print(results)
"""# Interactive Search Interface"""
import gradio as gr
from sklearn.metrics.pairwise import cosine_similarity
from PIL import Image
import requests
import pandas as pd
from sentence_transformers import SentenceTransformer
# تحميل نموذج Sentence Transformer لاستخراج المتجهات
embedding_model = SentenceTransformer('all-MiniLM-L6-v2')
# إعداد البيانات - هذه مجرد عينة من البيانات
data = {
"hotel_name": ["Hotel A", "Hotel B", "Hotel C"],
"hotel_description": [
"Beautiful hotel with sea view",
"Luxury hotel with great amenities",
"Affordable hotel with excellent location"
],
"summary": [
"Sea view and modern rooms.",
"Luxurious and spacious.",
"Economical and well-located."
],
"locality": ["London", "Paris", "New York"],
"hotel_image": [
"https://example.com/hotelA.jpg", # استبدل هذه الروابط بروابط صور حقيقية
"https://example.com/hotelB.jpg",
"https://example.com/hotelC.jpg"
]
}
summary_df = pd.DataFrame(data)
summary_df["embedding"] = summary_df["hotel_description"].apply(lambda x: embedding_model.encode(x))
def search(query):
query_embedding = embedding_model.encode(query)
summary_df["similarity"] = summary_df["embedding"].apply(lambda x: cosine_similarity([query_embedding], [x])[0][0])
results = summary_df.sort_values("similarity", ascending=False).head(3) # نعرض أفضل 3 نتائج
output = []
for index, row in results.iterrows():
hotel_info = f"**Hotel Name:** {row['hotel_name']}\n" \
f"**Description:** {row['hotel_description']}\n" \
f"**Summary:** {row['summary']}\n" \
f"**Locality:** {row['locality']}\n" \
f"**Similarity Score:** {row['similarity']:.2f}\n"
image_url = row['hotel_image']
image = Image.open(requests.get(image_url, stream=True).raw).resize((200, 200))
output.append((hotel_info, image))
return output
def gradio_interface(query):
results = search(query)
return results
# Create a Gradio interface
# Update to use gr.Textbox for both inputs and outputs
demo = gr.Interface(
fn=search_gradio,
inputs=gr.Textbox(lines=2, label="Enter your query"), # Use gr.Textbox directly
outputs=gr.Textbox(label="Search Results") # Use gr.Textbox directly
)
demo.launch() |