Spaces:
Running
Running
| import streamlit as st | |
| import pandas as pd | |
| import seaborn as sns | |
| import matplotlib.pyplot as plt | |
| import google.generativeai as genai | |
| import numpy as np | |
| # Set up Gemini API (replace with your actual API key) | |
| genai.configure(api_key='AIzaSyBDeJo3pioFL92ErFTtmRBmWt5diryp0E0') | |
| def load_and_preprocess_data(file): | |
| data = pd.read_csv(file) | |
| emotion_columns = ['angry', 'disgust', 'fear', 'happy', 'sad', 'surprise', 'neutral'] | |
| for col in emotion_columns: | |
| data[col] = pd.to_numeric(data[col], errors='coerce') | |
| data = data.dropna() | |
| return data | |
| def calculate_emotion_stats(data, emotion): | |
| return { | |
| 'mean': data[emotion].mean(), | |
| 'median': data[emotion].median(), | |
| 'std': data[emotion].std() | |
| } | |
| import streamlit as st | |
| import pandas as pd | |
| import seaborn as sns | |
| import matplotlib.pyplot as plt | |
| import numpy as np | |
| def calculate_emotion_percentages(data): | |
| emotion_columns = ['angry', 'disgust', 'fear', 'happy', 'sad', 'surprise', 'neutral'] | |
| total = data[emotion_columns].sum().sum() | |
| percentages = (data[emotion_columns].sum() / total * 100).round(2) | |
| return percentages | |
| def visualize_all_emotions_pie_chart(data_dict, selected_movies): | |
| emotion_percentages = {} | |
| for movie in selected_movies: | |
| fig, ax = plt.subplots(figsize=(10, 6)) | |
| percentages = calculate_emotion_percentages(data_dict[movie]) | |
| ax.pie(percentages, labels=percentages.index, autopct='%1.1f%%', startangle=90) | |
| ax.set_title(f"Distribution of Emotions in {movie}") | |
| st.pyplot(fig) | |
| emotion_percentages[movie] = percentages | |
| return emotion_percentages | |
| def visualize_comparison(data_dict, features, selected_movies): | |
| if not features: | |
| st.write("No features selected for visualization.") | |
| return None | |
| if 'all' in features: | |
| features = ['angry', 'disgust', 'fear', 'happy', 'sad', 'surprise', 'neutral'] | |
| return visualize_all_emotions_pie_chart(data_dict, selected_movies) | |
| else: | |
| visualize_specific_emotions(data_dict, features, selected_movies) | |
| return None | |
| import seaborn as sns | |
| import matplotlib.pyplot as plt | |
| def visualize_specific_emotions(data_dict, features, selected_movies): | |
| fig, ax = plt.subplots() | |
| for movie in selected_movies: | |
| for feature in features: | |
| # Check if the emotion exists in the data for this candidate | |
| if feature in data_dict[movie]: | |
| sns.kdeplot(data_dict[movie][feature], ax=ax, label=movie, shade=True) | |
| else: | |
| st.warning(f"'{feature}' data not found for {movie}. Skipping this feature.") | |
| plt.legend() | |
| st.pyplot(fig) | |
| if len(selected_movies) == 2: | |
| fig, ax = plt.subplots(figsize=(12, 10)) | |
| correlation = pd.concat([data_dict[movie][features] for movie in selected_movies], axis=1, keys=selected_movies).corr() | |
| sns.heatmap(correlation, annot=True, cmap='coolwarm', ax=ax) | |
| ax.set_title('Correlation Heatmap of Selected Features Between Movies') | |
| st.pyplot(fig) | |
| def visualize_single_candidate(data, emotions): | |
| if 'all' in emotions: | |
| # Pie chart for all emotions | |
| emotion_columns = ['angry', 'disgust', 'fear', 'happy', 'sad', 'surprise', 'neutral'] | |
| percentages = calculate_emotion_percentages(data) | |
| fig, ax = plt.subplots(figsize=(10, 6)) | |
| ax.pie(percentages, labels=percentages.index, autopct='%1.1f%%', startangle=90) | |
| ax.set_title(f"Distribution of All Emotions") | |
| st.pyplot(fig) | |
| return percentages.to_dict() | |
| else: | |
| # Distribution plot for specific emotions | |
| fig, ax = plt.subplots(figsize=(10, 6)) | |
| for emotion in emotions: | |
| sns.kdeplot(data[emotion], ax=ax, label=emotion, shade=True) | |
| ax.set_title(f"Distribution of Selected Emotions") | |
| ax.set_xlabel("Intensity") | |
| ax.set_ylabel("Density") | |
| ax.legend() | |
| st.pyplot(fig) | |
| return None | |
| def format_emotion_prompt(emotions, data_dict, selected_movies, emotion_percentages=None): | |
| prompt = """ | |
| You are an AI assistant specializing in movie emotion analysis. You have access to emotion data for the following movies: {MOVIES}, focusing on these emotions: {EMOTIONS}. | |
| {STATS} | |
| Based on this data: | |
| 1. Compare the overall levels of the specified emotions across the selected movies. Which movies exhibit more intense or frequent emotions? | |
| 2. Analyze the distribution of emotions in each selected movie. Are they evenly spread or concentrated at certain levels? | |
| 3. Discuss any significant differences in how these emotions are expressed across the selected movies. What might these differences suggest about the movies' content or style? | |
| 4. Consider the variability of emotions in each selected movie. Do some movies have more consistent levels of emotions, or do they fluctuate more? | |
| 5. Based on this emotion data, hypothesize about potential scenes or themes in each selected movie that might contribute to the observed emotion patterns. | |
| 6. How might the differences in emotions between these selected movies affect the overall viewer experience? | |
| Provide a detailed analysis addressing these points, using specific data references where relevant. Your analysis should offer insights into how these emotions are utilized in each selected movie and what this reveals about their emotional content and potential audience impact. | |
| """ | |
| stats = "" | |
| all_emotions = ['angry', 'disgust', 'fear', 'happy', 'sad', 'surprise', 'neutral'] | |
| for movie in selected_movies: | |
| stats += f"\n{movie}:\n" | |
| if emotion_percentages and movie in emotion_percentages: | |
| for emotion, percentage in emotion_percentages[movie].items(): | |
| stats += f"{emotion.capitalize()}: {percentage:.2f}%\n" | |
| else: | |
| emotions_to_analyze = all_emotions if 'all' in emotions else emotions | |
| for emotion in emotions_to_analyze: | |
| emotion_stats = calculate_emotion_stats(data_dict[movie], emotion) | |
| stats += f"{emotion.capitalize()} - Mean: {emotion_stats['mean']:.2f}, Median: {emotion_stats['median']:.2f}, Standard Deviation: {emotion_stats['std']:.2f}\n" | |
| emotions_display = "all emotions" if 'all' in emotions else ", ".join(emotions) | |
| return prompt.format(MOVIES=", ".join(selected_movies), EMOTIONS=emotions_display, STATS=stats) | |
| def generate_response(prompt, data_dict, emotions, selected_movies, emotion_percentages=None): | |
| model = genai.GenerativeModel('gemini-pro') | |
| analysis_prompt = format_emotion_prompt(emotions, data_dict, selected_movies, emotion_percentages) | |
| full_prompt = analysis_prompt + "\n\nUser query: " + prompt | |
| response = model.generate_content(full_prompt) | |
| if hasattr(response, 'candidates'): | |
| if response.candidates: | |
| content = response.candidates[0].content | |
| if hasattr(content, 'parts'): | |
| for part in content.parts: | |
| if hasattr(part, 'text'): | |
| return part.text | |
| return "Error: Unable to extract text from the response. Please check the API response structure." | |
| def main(): | |
| st.title("Multi-Movie Emotion Analysis Chat Interface") | |
| num_movies = st.number_input("How many movies would you like to compare?", min_value=1, max_value=10, value=1) | |
| data_dict = {} | |
| for i in range(num_movies): | |
| uploaded_file = st.file_uploader(f"Choose CSV file for Movie {i+1}", type="csv", key=f"movie_{i+1}") | |
| if uploaded_file is not None: | |
| data = load_and_preprocess_data(uploaded_file) | |
| data_dict[f"Movie {i+1}"] = data | |
| if len(data_dict) == num_movies: | |
| st.success("All files uploaded successfully. You can now start chatting!") | |
| st.subheader("Data Information") | |
| for movie, data in data_dict.items(): | |
| st.write(f"{movie} Columns:", data.columns.tolist()) | |
| emotions = ['angry', 'disgust', 'fear', 'happy', 'sad', 'surprise', 'neutral'] | |
| st.write("Available Emotions:", emotions) | |
| if "messages" not in st.session_state: | |
| st.session_state.messages = [] | |
| for message in st.session_state.messages: | |
| with st.chat_message(message["role"]): | |
| st.markdown(message["content"]) | |
| # Movie selection for each query | |
| selected_movies = st.multiselect("Select movies to compare :", list(data_dict.keys())) | |
| if not selected_movies: | |
| selected_movies = list(data_dict.keys()) | |
| if prompt := st.chat_input("What would you like to know about the movies' emotions?"): | |
| st.chat_message("user").markdown(prompt) | |
| st.session_state.messages.append({"role": "user", "content": prompt}) | |
| selected_emotions = [] | |
| if "all" in prompt.lower(): | |
| selected_emotions = ['all'] | |
| else: | |
| for emotion in emotions: | |
| if emotion in prompt.lower(): | |
| selected_emotions.append(emotion) | |
| if not selected_emotions: | |
| selected_emotions = ['all'] | |
| emotion_percentages = None | |
| with st.chat_message("assistant"): | |
| if len(selected_movies) == 1: | |
| # Single candidate scenario | |
| emotion_percentages = visualize_single_candidate(data_dict[selected_movies[0]], selected_emotions) | |
| elif any(keyword in prompt.lower() for keyword in ["graph", "compare", "visualize", "show"]): | |
| # Multiple candidates scenario | |
| emotion_percentages = visualize_comparison(data_dict, selected_emotions, selected_movies) | |
| response = generate_response(prompt, data_dict, selected_emotions, selected_movies, emotion_percentages) | |
| with st.chat_message("assistant"): | |
| st.markdown(response) | |
| st.session_state.messages.append({"role": "assistant", "content": response}) | |
| if st.checkbox("Show raw data"): | |
| for movie in selected_movies: | |
| st.subheader(f"{movie} Data") | |
| st.write(data_dict[movie]) | |
| if __name__ == "__main__": | |
| main() |