import pandas as pd import numpy as np import gradio as gr import os from sentence_transformers import SentenceTransformer, util import torch import re print("--- Film Öneri Sistemi Başlatılıyor ---") csv_file_name = "imdb-top-rated-movies-user-rated.csv" file_path = os.path.join(".", csv_file_name) if not os.path.exists(file_path): print(f"HATA: '{csv_file_name}' dosyası bulunamadı.") exit(1) try: df = pd.read_csv(file_path) print(f"'{csv_file_name}' başarıyla yüklendi. Toplam {len(df)} film bulundu.") except Exception as e: print(f"HATA: CSV dosyası yüklenirken hata oluştu: {e}") exit(1) df_filtered = df[['Title', 'IMDb Rating', 'Tags', 'Director', 'Stars', 'Votes', 'Description', 'Poster URL']].copy() df_filtered['Stars'].fillna('', inplace=True) df_filtered['Description'].fillna('', inplace=True) df_filtered['Poster URL'].fillna('', inplace=True) genre_mapping = { 'action': ['action', 'action epic', 'gun fu', 'one-person army action', 'car action', 'kung fu', 'martial arts', 'martial-arts'], 'adventure': ['adventure', 'adventure epic', 'desert adventure', 'animal adventure', 'space adventure', 'swashbuckler'], 'comedy': ['comedy', 'romantic comedy', 'buddy comedy', 'sitcom', 'black comedy', 'satire', 'spoof', 'parody', 'slapstick', 'screwball comedy', 'dark comedy', 'body swap comedy'], 'drama': ['drama', 'period drama', 'cop drama', 'legal drama', 'medical drama', 'teen drama', 'psychological drama', 'melodrama', 'historical drama', 'biography', 'romantic drama', 'showbiz drama', 'tragedy'], 'thriller': ['thriller', 'crime thriller', 'spy thriller', 'psychological thriller', 'mystery thriller', 'political thriller', 'conspiracy thriller', 'erotic thriller', 'cyber thriller', 'suspense'], 'sci-fi': ['sci-fi', 'space sci-fi', 'dystopian sci-fi', 'cyberpunk', 'alien invasion', 'mutant', 'robot', 'post-apocalyptic', 'time travel'], 'fantasy': ['fantasy', 'dark fantasy', 'sword & sorcery', 'fairy tale', 'epic fantasy'], 'horror': ['horror', 'slasher', 'supernatural horror', 'body horror', 'zombie', 'monster', 'vampire', 'werewolf', 'ghost'], 'mystery': ['mystery', 'suspense mystery', 'cozy mystery', 'whodunnit', 'detective', 'police procedural'], 'crime': ['crime', 'gangster', 'heist', 'mob', 'true crime'], 'romance': ['romance', 'romantic comedy', 'romantic drama'], 'animation': ['animation', 'adult animation', 'anime', 'computer animation', 'drawn animation', 'stop-motion animation'], 'family': ['family', 'kids'], 'western': ['western', 'classic western', 'neo-western'], 'war': ['war', 'war drama'], 'history': ['history', 'historical drama', 'biography'], 'music': ['music', 'musical', 'classic musical', 'concert'], 'documentary': ['documentary', 'docudrama', 'mockumentary'] } reverse_genre_map = {} for main_genre, sub_genres in genre_mapping.items(): for sub_genre in sub_genres: reverse_genre_map[sub_genre] = main_genre def map_to_main_genres(tag_list): main_genres = set() for tag in tag_list: if tag in reverse_genre_map: main_genres.add(reverse_genre_map[tag]) return list(main_genres) def clean_and_split(text_series): if pd.isna(text_series): return [] item = str(text_series) item = item.replace('"', '').replace("'", '').strip() item = item.replace('sci, fi', 'sci-fi') split_items = [s.strip().lower() for s in item.split(',') if s.strip()] return split_items df_filtered['Tags_cleaned_raw'] = df_filtered['Tags'].apply(clean_and_split) df_filtered['Director_cleaned'] = df_filtered['Director'].apply(clean_and_split) df_filtered['Stars_cleaned'] = df_filtered['Stars'].apply(clean_and_split) df_filtered['Tags_cleaned'] = df_filtered['Tags_cleaned_raw'].apply(map_to_main_genres) def convert_votes_to_numeric(votes_str): if isinstance(votes_str, str): votes_str = votes_str.replace(",", "") if 'K' in votes_str: return float(votes_str.replace('K', '')) * 1000 elif 'M' in votes_str: return float(votes_str.replace('M', '')) * 1_000_000 try: return float(votes_str) except ValueError: return np.nan df_filtered['Votes_numeric'] = df_filtered['Votes'].apply(convert_votes_to_numeric) df_filtered.drop('Votes', axis=1, inplace=True) df_filtered.dropna(subset=['Votes_numeric'], inplace=True) df_filtered['Combined_Text'] = df_filtered['Title'] + ". " + \ df_filtered['Description'] + ". " + \ df_filtered['Tags_cleaned'].apply(lambda x: ", ".join(x)) + ". " + \ df_filtered['Director_cleaned'].apply(lambda x: ", ".join(x)) + ". " + \ df_filtered['Stars_cleaned'].apply(lambda x: ", ".join(x)) model_name = 'sentence-transformers/all-MiniLM-L6-v2' try: sentence_model = SentenceTransformer(model_name) print(f"'{model_name}' modeli başarıyla yüklendi.") except Exception as e: print(f"HATA: Sentence Transformer modeli yüklenirken hata oluştu: {e}") exit(1) embeddings_file_path = os.path.join(".", "film_embeddings.npy") if not os.path.exists(embeddings_file_path): print(f"HATA: '{embeddings_file_path}' dosyası bulunamadı.") exit(1) try: film_embeddings = torch.from_numpy(np.load(embeddings_file_path)) print("Film embedding'leri başarıyla yüklendi.") except Exception as e: print(f"HATA: film_embeddings.npy yüklenirken hata oluştu: {e}") exit(1) director_popularity = {} for index, row in df_filtered.iterrows(): for director in row['Director_cleaned']: director_popularity[director] = director_popularity.get(director, 0) + row['Votes_numeric'] star_popularity = {} for index, row in df_filtered.iterrows(): for star in row['Stars_cleaned']: star_popularity[star] = star_popularity.get(star, 0) + row['Votes_numeric'] all_tags = sorted(list(set([tag for sublist in df_filtered['Tags_cleaned'] for tag in sublist if tag in genre_mapping]))) all_directors = sorted(list(director_popularity.keys()), key=lambda d: director_popularity[d], reverse=True) all_stars = sorted(list(star_popularity.keys()), key=lambda s: star_popularity[s], reverse=True) def get_movie_recommendations(selected_tags, selected_directors, selected_stars, min_imdb_rating_slider, num_recommendations_slider, search_text=""): selected_tags_list = list(selected_tags) if selected_tags else [] selected_directors_list = list(selected_directors) if selected_directors else [] selected_stars_list = list(selected_stars) if selected_stars else [] recommendations_df = df_filtered.copy() recommendations_df = recommendations_df[recommendations_df['IMDb Rating'] >= min_imdb_rating_slider] if selected_tags_list: recommendations_df = recommendations_df[ recommendations_df['Tags_cleaned'].apply(lambda x: any(tag in x for tag in selected_tags_list)) ] if selected_directors_list: recommendations_df = recommendations_df[ recommendations_df['Director_cleaned'].apply(lambda x: any(director in x for director in selected_directors_list)) ] if selected_stars_list: recommendations_df = recommendations_df[ recommendations_df['Stars_cleaned'].apply(lambda x: any(star in x for star in selected_stars_list)) ] if search_text and len(recommendations_df) > 0: try: query_embedding = sentence_model.encode(search_text, convert_to_tensor=True) filtered_indices = recommendations_df.index.tolist() current_film_embeddings = film_embeddings[filtered_indices] cosine_scores = util.cos_sim(query_embedding, current_film_embeddings)[0] recommendations_df['Similarity_Score'] = cosine_scores.cpu().numpy() recommendations_df = recommendations_df.sort_values( by=['Similarity_Score', 'IMDb Rating', 'Votes_numeric'], ascending=[False, False, False] ).reset_index(drop=True) except Exception as e: return "Benzerlik hesaplanırken bir hata oluştu." if not search_text: recommendations_df = recommendations_df.sort_values( by=['IMDb Rating', 'Votes_numeric'], ascending=[False, False] ).reset_index(drop=True) top_recommendations = recommendations_df.head(num_recommendations_slider) if top_recommendations.empty: return """
🎬

Sonuç Bulunamadı

Seçtiğiniz kriterlere uygun film bulunamadı. Filtreleri değiştirerek tekrar deneyin.

""" else: html_output = "" for idx, row in top_recommendations.iterrows(): directors_str = ", ".join([d.title() for d in row['Director_cleaned']]) stars_str = ", ".join([s.title() for s in row['Stars_cleaned']]) tags_str = ", ".join([t.title() for t in row['Tags_cleaned']]) similarity_info = "" if 'Similarity_Score' in row and search_text: sim_percentage = int(row['Similarity_Score'] * 100) similarity_info = f"""
🎯 Eşleşme: %{sim_percentage}
""" rating_color = "#4ade80" if row['IMDb Rating'] >= 8.0 else "#fbbf24" if row['IMDb Rating'] >= 7.5 else "#fb923c" poster_html = f"""
🎬
{row['Title'][:40]}... Poster Yok
""" html_output += f"""
{poster_html}

{row['Title']}

⭐ {row['IMDb Rating']:.1f}
{similarity_info}
🗳️ {int(row['Votes_numeric']):,} Oy
🎬 Yönetmen: {directors_str if directors_str else 'Bilinmiyor'}
⭐ Oyuncular: {stars_str if stars_str else 'Bilinmiyor'}
{''.join([f'{tag.title()}' for tag in row['Tags_cleaned']])}
""" return html_output with gr.Blocks(theme=gr.themes.Soft(), css=""" .gradio-container { max-width: 1200px !important; font-family: 'Segoe UI', sans-serif; } h1 { color: #f39c12; text-align: center; } h3 { color: #eee; } .gr-button.gr-button-primary { background-color: #f39c12 !important; border-color: #f39c12 !important; } .gr-button.gr-button-primary:hover { background-color: #e67e22 !important; border-color: #e67e22 !important; } .gr-checkbox-group label { color: #ccc; } .gr-dropdown, .gr-slider, .gr-textbox { background-color: #2c2c2c; color: #eee; border-color: #555; } .gr-dropdown-item { color: #eee; } .gr-checkbox-group input[type='checkbox']:checked + label { background-color: #f39c12 !important; border-color: #f39c12 !important; color: #1a1a1a !important; } .gr-checkbox-group input[type='checkbox'] + label { background-color: #333; color: #eee; border: 1px solid #555; } .gr-checkbox-group input[type='checkbox'] + label:hover { background-color: #444; } .gr-dropdown-item.selected { background-color: #f39c12 !important; color: #1a1a1a !important; } .gr-dropdown-item:hover { background-color: #e67e22 !important; color: #1a1a1a !important; } input[type="range"]::-webkit-slider-thumb { background-color: #f39c12 !important; } input[type="range"]::-moz-range-thumb { background-color: #f39c12 !important; } """) as demo: gr.Markdown( """ # 🎬 Film Öneri Sistemi Favori film özelliklerinizi seçin, yüksek IMDb puanına sahip filmleri keşfedin! İstediğiniz bir film veya konu hakkında yazın, benzerlerini de bulalım. """ ) with gr.Row(): with gr.Column(scale=2): gr.Markdown("### Önerilen Filmler:") output_html = gr.HTML(label="Önerileriniz burada listelenecektir.", value="

Henüz bir öneri yapılmadı. Özellikleri seçip butona tıklayın!

") with gr.Row(): with gr.Column(scale=1): gr.Markdown("### Film Özelliklerini Seçin:") tags_input = gr.CheckboxGroup( label="Film Türleri", choices=all_tags, value=['action', 'drama'], interactive=True ) directors_input = gr.Dropdown( label="Yönetmenler", choices=all_directors, multiselect=True, allow_custom_value=False, interactive=True ) stars_input = gr.Dropdown( label="Oyuncular", choices=all_stars, multiselect=True, allow_custom_value=False, interactive=True ) min_imdb_rating_slider = gr.Slider( minimum=df_filtered['IMDb Rating'].min(), maximum=df_filtered['IMDb Rating'].max(), step=0.1, value=7.6, label="Minimum IMDb Puanı" ) num_recommendations_slider = gr.Slider( minimum=1, maximum=20, step=1, value=10, label="Öneri Sayısı" ) search_text_input = gr.Textbox( label="Film Adı veya Konu Hakkında Ara (NLP Tabanlı Benzerlik)", placeholder="Örneğin: Batman, uzay filmi, zamanda yolculuk..." ) recommend_btn = gr.Button("🚀 Film Önerilerini Getir", variant="primary", size="lg") recommend_btn.click( fn=get_movie_recommendations, inputs=[tags_input, directors_input, stars_input, min_imdb_rating_slider, num_recommendations_slider, search_text_input], outputs=output_html ) gr.Examples( examples=[ [['action'], [], [], 7.6, 5, ""], [['comedy', 'drama'], [], [], 7.8, 3, ""], [[], ['christopher nolan'], [], 8.0, 5, ""], [[], [], ['leonardo dicaprio'], 7.8, 3, ""], [[], [], [], 8.0, 5, "kahramanlık ve bilim kurgu"], [['action', 'sci-fi'], [], [], 7.8, 5, "uzaylı istilası ve kaçış"], ], inputs=[tags_input, directors_input, stars_input, min_imdb_rating_slider, num_recommendations_slider, search_text_input], outputs=output_html, fn=get_movie_recommendations, label="Örnek Önerileri Deneyin" ) gr.Markdown( """ --- ### ℹ️ Nasıl Kullanılır? 1. **Film Türleri, Yönetmenler ve Oyuncular** bölümlerinden istediğiniz filtreleri seçin (birden fazla seçim yapabilirsiniz). 2. **Minimum IMDb Puanı** ve **Öneri Sayısı** çubuklarını ayarlayın. 3. İsterseniz **"Film Adı veya Konu Hakkında Ara"** kutucuğuna bir film adı, konu veya anahtar kelime yazın. 4. **"🚀 Film Önerilerini Getir"** butonuna tıklayın. 5. Öneriler üst panelde görünecektir! """ ) demo.launch(share=True) print("Gradio Web Arayüzü Başlatıldı.")