import gradio as gr from transformers import pipeline import numpy as np from gradio_client import Client import matplotlib.pyplot as plt from lime.lime_text import LimeTextExplainer from datetime import datetime import sqlite3 import logging DB_NAME = "mental_health_analysis.db" # --- Database Setup --- def create_database(): conn = sqlite3.connect(DB_NAME) cursor = conn.cursor() cursor.execute(""" CREATE TABLE IF NOT EXISTS mental_health_analysis ( id INTEGER PRIMARY KEY AUTOINCREMENT, text TEXT NOT NULL, sentiment TEXT, sentiment_confidence REAL, disorder TEXT, disorder_confidence REAL, risk_level TEXT, recommendations TEXT, date TEXT ) """) conn.commit() conn.close() # --- Insert Results into Database --- def insert_into_database(text, sentiment, sentiment_confidence, disorder, disorder_confidence, risk_level, recommendations, date): conn = sqlite3.connect(DB_NAME) cursor = conn.cursor() # Convert list to string for recommendations recommendations_str = "; ".join(recommendations) cursor.execute(""" INSERT INTO mental_health_analysis ( text, sentiment, sentiment_confidence, disorder, disorder_confidence, risk_level, recommendations, date ) VALUES (?, ?, ?, ?, ?, ?, ?, ?) """, ( text, sentiment, sentiment_confidence, disorder, disorder_confidence, risk_level, recommendations_str, date )) conn.commit() conn.close() create_database() class MentalHealthChatbot: def __init__(self): self.sentiment_client = Client("hasanmustafa0503/sentiment") self.disorder_client = Client("hasanmustafa0503/disoreder-api") # Label mappings self.label_mapping = { "LABEL_0": "ADHD", "LABEL_1": "BPD", "LABEL_2": "OCD", "LABEL_3": "PTSD", "LABEL_4": "Anxiety", "LABEL_5": "Autism", "LABEL_6": "Bipolar", "LABEL_7": "Depression", "LABEL_8": "Eating Disorders", "LABEL_9": "Health", "LABEL_10": "Mental Illness", "LABEL_11": "Schizophrenia", "LABEL_12": "Suicide Watch" } self.sentiment_mapping = { "POS": "Positive", "NEG": "Negative", "NEU": "Neutral" } self.exercise_recommendations = { # Exercise recommendations data as defined in the original code } # Initialize the LIME explainer self.explainer = LimeTextExplainer(class_names=list(self.sentiment_mapping.values()) + list(self.label_mapping.values())) def get_sentiment(self, text): results = self.sentiment_client.predict(text=text, api_name="/predict") if results and isinstance(results, list): label = results[0]["label"] confidence_score = results[0]["score"]*100 sentiment = self.sentiment_mapping.get(label, "Unknown") return sentiment, confidence_score return "Unknown", 0 def get_disorder(self, text, threshold=35): results = self.disorder_client.predict(text=text, api_name="/predict") if results and isinstance(results, list): best_result = results[0] disorder_confidence = best_result["score"] * 100 if disorder_confidence > threshold: disorder_label = self.label_mapping.get(best_result["label"], "Unknown Disorder") if disorder_confidence < 50: risk_level = "Low Risk" elif 50 <= disorder_confidence <= 75: risk_level = "Moderate Risk" else: risk_level = "High Risk" return disorder_label, disorder_confidence, risk_level return "No significant disorder detected", 0.0, "No Risk" def predict_fn(self, texts): sentiment_probs = [] disorder_probs = [] sentiment_labels = [] disorder_labels = [] for text in texts: sentiment, sentiment_confidence = self.get_sentiment(text) sentiment_probs.append([sentiment_confidence / 100]) sentiment_labels.append(sentiment) disorder_label, disorder_confidence, risk_level = self.get_disorder(text) disorder_probs.append([disorder_confidence / 100]) disorder_labels.append(disorder_label) sentiment_probs = np.array(sentiment_probs) disorder_probs = np.array(disorder_probs) result = np.hstack([sentiment_probs, disorder_probs]) return result, sentiment_labels, disorder_labels def lime_predict_fn(self, texts): result, _, _ = self.predict_fn(texts) return result def explain_text(self, text): explanation = self.explainer.explain_instance(text, self.lime_predict_fn, num_features=5, num_samples=25) explanation.as_pyplot_figure() # Display the plot plt.show() explanation_str = "The model's prediction is influenced by the following factors: " explanation_str += "; ".join([f'"{feature}" contributes with a weight of {weight:.4f}' for feature, weight in explanation.as_list()]) + "." return explanation_str def get_recommendations(self, condition, risk_level): exercise_recommendations = { "Depression": { "High Risk": ["Try 10 minutes of deep breathing.", "Go for a 15-minute walk in nature.", "Practice guided meditation."], "Moderate Risk": ["Write down 3 things you’re grateful for.", "Do light stretching or yoga for 10 minutes.", "Listen to calming music."], "Low Risk": ["Engage in a hobby you enjoy.", "Call a friend and have a short chat.", "Do a short 5-minute mindfulness exercise."] }, "Anxiety": { "High Risk": ["Try progressive muscle relaxation.", "Use the 4-7-8 breathing technique.", "Write down your thoughts to clear your mind."], "Moderate Risk": ["Listen to nature sounds or white noise.", "Take a 15-minute break from screens.", "Try a short visualization exercise."], "Low Risk": ["Practice slow, deep breathing for 5 minutes.", "Drink herbal tea and relax.", "Read a book for 10 minutes."] }, "Bipolar": { "High Risk": ["Engage in grounding techniques like 5-4-3-2-1.", "Try slow-paced walking in a quiet area.", "Listen to calm instrumental music."], "Moderate Risk": ["Do a 10-minute gentle yoga session.", "Keep a mood journal for self-awareness.", "Practice self-affirmations."], "Low Risk": ["Engage in light exercise like jogging.", "Practice mindful eating for a meal.", "Do deep breathing exercises."] }, "OCD": { "High Risk": ["Use exposure-response prevention techniques.", "Try 5 minutes of guided meditation.", "Write down intrusive thoughts and challenge them."], "Moderate Risk": ["Take a short break from triggers.", "Practice progressive relaxation.", "Engage in a calming activity like drawing."], "Low Risk": ["Practice deep breathing with slow exhales.", "Listen to soft music and relax.", "Try focusing on one simple task at a time."] }, "PTSD": { "High Risk": ["Try grounding techniques (hold an object, describe it).", "Do 4-7-8 breathing for relaxation.", "Write in a trauma journal."], "Moderate Risk": ["Practice mindfulness for 5 minutes.", "Engage in slow, rhythmic movement (walking, stretching).", "Listen to soothing music."], "Low Risk": ["Try positive visualization techniques.", "Engage in light exercise or stretching.", "Spend time in a quiet, safe space."] }, "Suicide Watch": { "High Risk": ["Immediately reach out to a mental health professional.", "Call a trusted friend or family member.", "Try a grounding exercise like cold water on hands."], "Moderate Risk": ["Write a letter to your future self.", "Listen to uplifting music.", "Practice self-care (take a bath, make tea, etc.)."], "Low Risk": ["Watch a motivational video.", "Write down your emotions in a journal.", "Spend time with loved ones."] }, "ADHD": { "High Risk": ["Try structured routines for the day.", "Use a timer for focus sessions.", "Engage in short bursts of physical activity."], "Moderate Risk": ["Do a quick exercise routine (jumping jacks, stretches).", "Use fidget toys to channel energy.", "Try meditation with background music."], "Low Risk": ["Practice deep breathing.", "Listen to classical or instrumental music.", "Organize your workspace."] }, "BPD": { "High Risk": ["Try dialectical behavior therapy (DBT) techniques.", "Practice mindfulness.", "Use a weighted blanket for comfort."], "Moderate Risk": ["Write down emotions and analyze them.", "Engage in creative activities like painting.", "Listen to calming podcasts."], "Low Risk": ["Watch a lighthearted movie.", "Do breathing exercises.", "Call a friend for a short chat."] }, "Autism": { "High Risk": ["Engage in deep-pressure therapy (weighted blanket).", "Use noise-canceling headphones.", "Try sensory-friendly relaxation techniques."], "Moderate Risk": ["Do repetitive physical activities like rocking.", "Practice structured breathing exercises.", "Engage in puzzles or memory games."], "Low Risk": ["Spend time in a quiet space.", "Listen to soft instrumental music.", "Follow a structured schedule."] }, "Schizophrenia": { "High Risk": ["Seek immediate support from a trusted person.", "Try simple grounding exercises.", "Use distraction techniques like puzzles."], "Moderate Risk": ["Engage in light physical activity.", "Listen to calming sounds or music.", "Write thoughts in a journal."], "Low Risk": ["Read a familiar book.", "Do a 5-minute breathing exercise.", "Try progressive muscle relaxation."] }, "Eating Disorders": { "High Risk": ["Seek professional help immediately.", "Try self-affirmations.", "Practice intuitive eating (listen to body cues)."], "Moderate Risk": ["Engage in mindful eating.", "Write down your emotions before meals.", "Do light stretching after meals."], "Low Risk": ["Try a gentle walk after eating.", "Listen to calming music.", "Write a gratitude journal about your body."] }, "Mental Health": { "High Risk": ["Reach out to a mental health professional.", "Engage in deep relaxation techniques.", "Talk to a support group."], "Moderate Risk": ["Write in a daily journal.", "Practice guided meditation.", "Do light physical activities like walking."], "Low Risk": ["Try deep breathing exercises.", "Watch an uplifting video.", "Call a friend for a chat."] } } if condition in exercise_recommendations: if risk_level in exercise_recommendations[condition]: return exercise_recommendations[condition][risk_level] return ["No specific recommendations available."] bot = MentalHealthChatbot() def analyze_text(text): if not text.strip(): return "No input provided.", "", "", "", "", "" disorder, disorder_conf, risk = bot.get_disorder(text) sentiment, sentiment_conf = bot.get_sentiment(text) if risk == "High Risk": logging.info("🚨 High risk detected! Alert triggered.") alert_msg = "✔ 🚨 Alert Notification Triggered: High risk detected!" else: alert_msg = "✔ Risk is not high. No alert triggered." recs = bot.get_recommendations(disorder, risk) lime_explanation = bot.explain_text(text) timestamp = datetime.now().strftime("%Y-%m-%d %H:%M:%S") # Insert data into DB insert_into_database( text, sentiment, sentiment_conf, disorder, disorder_conf, risk, recs, timestamp ) return ( f"{sentiment} ({sentiment_conf:.2f}%)", f"{disorder} ({disorder_conf:.2f}%)", risk, "; ".join(recs), lime_explanation, alert_msg ) # Gradio interface gr.Interface( fn=analyze_text, inputs=gr.Textbox(lines=6, label="Describe how you're feeling..."), outputs=[ gr.Text(label="Sentiment"), gr.Text(label="Disorder"), gr.Text(label="Risk Level"), gr.Text(label="Recommendations"), gr.Text(label="LIME Explanation"), gr.Text(label="Alert Message") ], title="🧠 Mental Health Analysis Assistant", description="This tool uses AI to detect mental health conditions based on your input and suggest possible actions." ).launch()