Added weighted metadata to the model

data/actors_set.csv

@@ -1150,6 +1150,7 @@ Joe Cappelletti
 Kenshô Ono
 Reina Ueda
 Jun'ichi Suwabe
+Toa Yukinari
 Kôhei Kiyasu
 Kenji Utsumi
 Ayumu Murase
@@ -1162,15 +1163,14 @@ John Rafter Lee
 Shigeru Matsuzaki
 Akiko Nakamura
 Toshiko Fujita
-Keiko Han
 Aya Suzaki
 Atsushi Tamaru
 Yûki Kaneko
 Lilas Ikuta
 Ano
-Erin Yvette
-Stephen Fu
-Grace Lu
+Yumi Kawai
+Mizuki Yoshida
+Yôichirô Saitô
 Kengo Kawanishi
 Shun Oguri
 Haruma Miura
@@ -1276,9 +1276,6 @@ Kirby Morrow
 Shôta Sometani
 Tetsurô Sagawa
 Gorô Naya
-Debora Rabbai
-Greg Wolfe
-Elisa Wain
 Shôtarô Morikubo
 Ryuji Aigase
 Takehito Koyasu
@@ -2372,7 +2369,7 @@ Akshay Kumar
 Sonu Sood
 Forest Whitaker
 Gong Chan-shik
-Won Tae Min
+Won Tae-min
 Do Woo
 Dominic Cooper
 Tyrone Power
@@ -3119,9 +3116,6 @@ Joe Rogan
 Kochi
 Avijit Halder
 Shanti Das
-Kerem Bürsin
-Dan Babic
-Adrianna Costa
 Dan Cohen
 Louise Dueno
 Nell Hardie
@@ -3136,9 +3130,6 @@ Joel Sartore
 Shndar Noradin Ali
 Ahmed Noradin Ali
 Noradin Khalaf Ali
-Jerry Springer
-Angel Anes
-Maria Gara
 Peter Andre
 Dominic Applewhite
 Bruce Dickinson
@@ -3372,10 +3363,6 @@ Eloise Eonnet
 Hadrian Dagannaud-Brouard
 Caitlin Zerra Rose
 Roy Wood Jr.
-Shirley Alvarez
-Melissa Arroyo
-Sarah Ashworth
-Joe Francis
 Cory Lane
 Taryn Carter
 Jacqueline Lovell
@@ -3386,9 +3373,6 @@ Vitaly S. Alexius
 Jeff Dunham
 Joel Johansson
 Jason 'Wee Man' Acuña
-Shah Rukh Khan
-Saif Ali Khan
-Rajkumar Hirani
 Derren Brown
 Richard Pope
 Simon Tcherniak

data/directors_set.csv

@@ -456,7 +456,6 @@ Satoshi Nishimura
 Susumu Mitsunaka
 Naoki Miyahara
 Osamu Dezaki
-Jun'ichi Satô
 Tomoyuki Kurokawa
 Kiyotaka Oshiyama
 Kenji Kamiyama
@@ -480,7 +479,6 @@ Yûsuke Yamamoto
 Takayuki Hirao
 Gorô Miyazaki
 Sunao Katabuchi
-Toshiya Shinohara
 Mayumi Nishimoto
 Hiroaki Akagi
 Taichi Ishidate
@@ -509,12 +507,10 @@ Shinji Itadaki
 Junji Shimizu
 Naoko Kusumi
 Mizuho Nishikubo
-See production info at IMDbPro
 Yasunao Aoki
 Nana Harada
 Hiroshi Haraguchi
 Keisuke Inoue
-Iku Suzuki
 Hirotsugu Kawasaki
 Norihiko Sutô
 Atsushi Takeuchi
@@ -856,7 +852,6 @@ David McMahon
 Brian Flemming
 Gaurav Jani
 Gabriela Cowperthwaite
-Jillian Schlesinger
 Chuan Lu
 Raoul Peck
 Joshua Oppenheimer
@@ -1356,7 +1351,6 @@ Anders Østergaard
 Brett Harvey
 Zana Briski
 Ross Kauffman
-Steve Meyer
 Michael Rossato-Bennett
 Adnan Al-Kaissy
 Lou Albano
@@ -1364,10 +1358,6 @@ Muhammad Ali
 Franny Armstrong
 John Scheinfeld
 Karzan Kardozi
-Panagiotis Tsartsianidis
-Jerry Springer
-Angel Anes
-Maria Gara
 Chris Howe
 Sam Wrench
 Bradley Cooper
@@ -1476,8 +1466,6 @@ Etan Cohen
 Johnny Knoxville
 Steve-O
 Chris Pontius
-Scott Cope
-Joe Francis
 Jason Williams
 Glenn Weiss
 Nick Sweeney
@@ -1490,8 +1478,6 @@ Richard A. Preuss
 Joel Johansson
 Bill Chapman
 Jason 'Wee Man' Acuña
-Shah Rukh Khan
-Saif Ali Khan
 Troy Miller
 Spencer Davis Gray
 Ricardo Benítez Garrido

data/genres_set.csv

@@ -51,6 +51,7 @@ Tragedy
 Slapstick
 Globetrotting Adventure
 Spy
+Political Thriller
 Psychological Thriller
 Caper
 Docudrama
@@ -88,7 +89,6 @@ Hand-Drawn Animation
 Pop Musical
 Buddy Cop
 Survival
-Political Thriller
 Teen Fantasy
 Spaghetti Western
 Western

src/app/how to run.txt

@@ -0,0 +1,4 @@
+cd src folder
+python app/main.py
+
+example C:\UAM-Repos\PJN-PROJEKT\src>python app/main.py

src/app/templates/results.html

@@ -14,10 +14,14 @@
             {% for recommendation in recommendations %}
                 <li class="list-group-item">
                     <h5><strong>{{ recommendation.title }}</strong></h5>
-                    <p><strong>Rating:</strong> {{ recommendation.rating }}</p>
+                    <p><strong>Release Date:</strong> {{ recommendation.release_date }}</p>
+                    <p><strong>Director:</strong> {{ recommendation.directors }}</p>
+                    <p><strong>Cast:</strong> {{ recommendation.stars }}</p>
                     <p><strong>Genres:</strong> {{ recommendation.genres }}</p>
                     <p><strong>Description:</strong> {{ recommendation.description }}</p>
                     <p><strong>Duration:</strong> {{ recommendation.duration_minutes }} minutes</p>
+                    <p><strong>Rating:</strong> {{ recommendation.rating }}</p>
+
                     <p>
                         <strong>More Info:</strong> 
                         <a href="{{ recommendation.url }}" target="_blank" class="link-primary">View Details</a>

src/model/recommend_movies.py

@@ -1,42 +1,100 @@
+import sys
+import os
+sys.path.append(os.path.abspath(os.path.join(os.path.dirname(__file__), '..')))
 import pandas as pd
 import pickle
 from transformers import BertTokenizer, BertModel
 from sentence_transformers import SentenceTransformer
 from sklearn.metrics.pairwise import cosine_similarity
+from process_input.analyze_input import analyze_users_input
 
 
 def load_embeddings(embeddings_path):
-
     with open(embeddings_path, 'rb') as f:
         embeddings = pickle.load(f)
     return embeddings
 
 
+def enhance_user_input(user_input):
+    analyzed_data = analyze_users_input(user_input)
+    parts = []
+
+    if analyzed_data['actors']:
+        parts.append(f"featuring {' and '.join(analyzed_data['actors'])}")
+    if analyzed_data['directors']:
+        parts.append(f"directed by {' and '.join(analyzed_data['directors'])}")
+    if analyzed_data['genres']:
+        parts.append(f"in the genre of {' and '.join(analyzed_data['genres'])}")
+    if analyzed_data['years']:
+        start, end = analyzed_data['years']
+        parts.append(f"released between {start} and {end}")
+    if analyzed_data['duration']:
+        parts.append(f"with a runtime under {analyzed_data['duration']} minutes")
+
+    print(f"Enhanced Input: {user_input} -> {', '.join(parts)}")
+
+    return " ".join(parts) or user_input
+
+
 def generate_user_embedding(user_input, model_type, model, tokenizer=None):
+    enhanced_input = enhance_user_input(user_input)
+    print(f"Enhanced Input for Embedding: {enhanced_input}")
+
     if model_type == 'bert':
         if tokenizer is None:
             raise ValueError("Tokenizer is required for BERT model.")
-        inputs = tokenizer(user_input, return_tensors='pt', max_length=512, truncation=True, padding='max_length')
+        inputs = tokenizer(enhanced_input, return_tensors='pt', max_length=512, truncation=True, padding='max_length')
         outputs = model(**inputs)
         return outputs.last_hidden_state[:, 0, :].squeeze(0).detach().numpy()
     elif model_type in ['sentence-transformer', 'sentence-bert']:
-        return model.encode(user_input, convert_to_tensor=False)
+        return model.encode(enhanced_input, convert_to_tensor=False)
     else:
         raise ValueError("Invalid model type. Choose 'bert', 'sentence-transformer', or 'sentence-bert'.")
 
 
 def recommend_movies(user_input, df, embeddings, model_type, model, tokenizer=None):
-    
     user_embedding = generate_user_embedding(user_input, model_type, model, tokenizer)
 
     similarities = cosine_similarity([user_embedding], embeddings).flatten()
 
-    df['similarity'] = similarities
+    analyzed_data = analyze_users_input(user_input)
+
+    weights = pd.Series(1.0, index=df.index) 
+
+    if analyzed_data['actors']:
+        for actor in analyzed_data['actors']:
+            weights += df['stars'].str.contains(actor, case=False, na=False) * 8
+
+    if analyzed_data['directors']:
+        for director in analyzed_data['directors']:
+            weights += df['directors'].str.contains(director, case=False, na=False) * 8
+
+    if analyzed_data['genres']:
+        for genre in analyzed_data['genres']:
+            weights += df['genres'].str.contains(genre, case=False, na=False) * 8  
+
+    if analyzed_data['years']:
+        start_year, end_year = map(int, analyzed_data['years'])
+        df_years = pd.to_numeric(df['release_date'], errors='coerce').fillna(0).astype(int)
+        weights += ((df_years >= start_year) & (df_years <= end_year)) * 8  
+
+    if analyzed_data['duration']:
+        weights += (df['duration_minutes'] <= analyzed_data['duration']) * 4
+
+    final_scores = similarities * weights
+
+    df['similarity'] = final_scores
 
     recommendations = df.sort_values(by='similarity', ascending=False).head(10)
 
     recommendations['genres'] = recommendations['genres'].apply(
         lambda x: ', '.join(eval(x)) if isinstance(x, str) and x.startswith('[') else x
     )
+    recommendations['directors'] = recommendations['directors'].apply(
+        lambda x: ', '.join(eval(x)) if isinstance(x, str) and x.startswith('[') else x
+    )
+    recommendations['stars'] = recommendations['stars'].apply(
+        lambda x: ', '.join(eval(x)) if isinstance(x, str) and x.startswith('[') else x
+    )
 
-    return recommendations[['title', 'rating', 'genres', 'description', 'duration_minutes', 'url', 'similarity']]
+    return recommendations[['title', 'rating', 'genres', 'release_date', 'directors', 'description', 'duration_minutes', 'stars', 'url', 'similarity']]

src/process_data/make_actors_set.py

@@ -30,7 +30,7 @@ def prepare_set_with_regex(data_path, save_path=None):
                 actors_set.append(actor)
 
     if save_path is not None:
-        with open(save_path, "w", newline="") as file:
+        with open(save_path, "w", newline="", encoding="utf-8") as file:
             writer = csv.writer(file)
 
             for actor in actors_set:
@@ -39,6 +39,7 @@ def prepare_set_with_regex(data_path, save_path=None):
     return
 
 
+
 def prepare_set_from_list(data_path, save_path=None):
     df = pd.read_csv(data_path, converters={
         'stars': ast.literal_eval,

src/process_data/make_directors_set.py

@@ -14,11 +14,13 @@ def prepare_set_from_list(data_path, save_path=None):
                 directors_set.append(director)
 
     if save_path is not None:
-        with open(save_path, "w", newline="") as file:
+        with open(save_path, "w", newline="", encoding="utf-8") as file:
             writer = csv.writer(file)
             for director in directors_set:
                 writer.writerow([director])
 
 
 data_path = "../../data/movies_data.csv"
-save_path = "../../data/directors_set.csv"
+save_path = "../../data/directors_set.csv"
+
+prepare_set_from_list(data_path, save_path)

src/process_data/make_genres_set.py

@@ -27,7 +27,7 @@ def prepare_set_with_regex(data_path, save_path=None):
                 genres_set.append(genre)
 
     if save_path is not None:
-        with open(save_path, "w", newline="") as file:
+        with open(save_path, "w", newline="", encoding="utf-8") as file:
             writer = csv.writer(file)
             for genre in genres_set:
                 writer.writerow([genre])
@@ -47,7 +47,7 @@ def prepare_set_from_list(data_path, save_path=None):
                 genres_set.append(genre)
 
     if save_path is not None:
-        with open(save_path, "w", newline="") as file:
+        with open(save_path, "w", newline="", encoding="utf-8") as file:
             writer = csv.writer(file)
             for genre in genres_set:
                 writer.writerow([genre])

src/process_data/preprocess_data.py

@@ -1,8 +1,7 @@
 import pandas as pd
 
 def convert_duration_to_minutes(duration):
-    
-    if not isinstance(duration, str): 
+    if not isinstance(duration, str):
         return None
     try:
         parts = duration.split()
@@ -10,39 +9,47 @@ def convert_duration_to_minutes(duration):
         minutes = int(parts[1][:-1]) if len(parts) > 1 and 'm' in parts[1] else 0
         return hours * 60 + minutes
     except (ValueError, IndexError):
-        return None 
+        return None
+
+
+def format_field(field):
+    if isinstance(field, list):
+        return ", ".join(field)
+    if isinstance(field, str) and field.startswith('['):
+        return ", ".join(eval(field))
+    return field
+
 
-    
 def load_and_preprocess(data_path, save_path=None):
     print("Loading data...")
-
     df = pd.read_csv(data_path)
 
     print("Preprocessing data...")
-    
+
     df['duration_minutes'] = df['duration'].apply(convert_duration_to_minutes)
 
     df = df[df['duration_minutes'].notnull() & (df['duration_minutes'] >= 60)]
 
     columns_to_parse = ['genres', 'directors', 'stars', 'keywords']
     for column in columns_to_parse:
-        df[column] = df[column].apply(eval)
-    
-    
+        df[column] = df[column].apply(eval).apply(format_field)
+
     df['description'] = df['description'].fillna('').astype(str)
     df['storyline'] = df['storyline'].fillna('').astype(str)
 
-    
     df['combined_text'] = (
-        df['description'] + " " + 
-        df['storyline'] + " " + 
-        df['keywords'].apply(lambda x: " ".join(x))
+        df['description'] + " " +
+        df['storyline'] + " " +
+        df['keywords'] + " " +
+        "featuring " + df['stars'] + " " +
+        "directed by " + df['directors'] + " " +
+        "in the genre of " + df['genres'] + " " +
+        "released in " + df['release_date'].astype(str) + " " +
+        "with a runtime of " + df['duration_minutes'].astype(str) + " minutes"
     )
 
-
     df = df.drop(columns=['storyline', 'keywords', 'duration'])
 
-
     print("Preprocessing complete.")
 
     if save_path:
@@ -51,10 +58,12 @@ def load_and_preprocess(data_path, save_path=None):
 
     return df
 
-data_path = "../../data/movies_data.csv" 
-save_path = "../../data/preprocessed_data.csv"  
+
+data_path = "../../data/movies_data.csv"
+save_path = "../../data/preprocessed_data.csv"
 
 df = load_and_preprocess(data_path, save_path)
 
+# Uncomment to preview the preprocessed data
 # print("\nPreprocessed Data Preview:")
-# print(df.head())
+# print(df.head())

src/process_input/analyze_input.py

@@ -1,22 +1,22 @@
 import spacy
 import pandas as pd
-from src.process_input.get_duration import get_movie_duration
-from src.process_input.get_genres import get_movie_genres
-from src.process_input.get_people import get_actor_or_director
-from src.process_input.get_years import get_movie_years
+from process_input.get_duration import get_movie_duration
+from process_input.get_genres import get_movie_genres
+from process_input.get_people import get_actor_or_director
+from process_input.get_years import get_movie_years
 
 
 def load_data():
     sets = {}
-    df = pd.read_csv('../../data/directors_set.csv', header=None)
+    df = pd.read_csv('../data/directors_set.csv', header=None)
     df.columns = ['Name']
     sets["directors"] = df['Name'].str.lower().tolist()
 
-    df = pd.read_csv('../../data/actors_set.csv', header=None)
+    df = pd.read_csv('../data/actors_set.csv', header=None)
     df.columns = ['Name']
     sets["actors"] = df['Name'].str.lower().tolist()
 
-    df = pd.read_csv('../../data/genres_set.csv', header=None)
+    df = pd.read_csv('../data/genres_set.csv', header=None)
     df.columns = ['Name']
     sets["genres"] = df['Name'].str.lower().tolist()