app_diff.py.diff

@app.route('/generate_sentiment', methods=['POST'])
def generate_sentiment():
    if "admin_id" not in session:
        flash("Access denied!", "danger")
        return redirect(url_for("login"))

    conn = get_db_connection()
    if not conn:
        flash("Database connection failed", "danger")
        return redirect(url_for("admin_dashboard"))
        
    cursor = conn.cursor()

    # 0. Clear old results
    try:
        cursor.execute("DELETE FROM sentiment_score")
        conn.commit()
    except Exception as e:
        print("Error clearing old sentiment results:", e)

    # 1. Fetch all comments
    cursor.execute("SELECT username, comment, discussion_topic FROM comment")
    comments = cursor.fetchall()

    if not comments:
        flash("No comments found for analysis", "warning")
        cursor.close()
        conn.close()
        return redirect(url_for("admin_dashboard"))

    # 2. Convert to DataFrame for easier processing
    df = pd.DataFrame(comments, columns=['username','comment','discussion_topic'])

    # Clean text
    df["comment_clean"] = (
        df["comment"].astype(str)
        .str.strip()
        .str.replace(r"\s+", " ", regex=True)
    )
    df = df[df["comment_clean"].str.len() > 2].copy()

    # 3. Load HuggingFace pipeline
    device = 0 if torch.cuda.is_available() else -1
    sent_pipe = pipeline("sentiment-analysis",
                         model="distilbert-base-uncased-finetuned-sst-2-english",
                         device=device,
                         return_all_scores=True)

    # 4. Run sentiment analysis
    pos_scores, neg_scores, sentiments, labels = [], [], [], []
    for comment in df['comment_clean'].tolist():
        result = sent_pipe(comment[:512])[0]
        scores = {r['label'].upper(): r['score'] for r in result}
        pos = scores.get('POSITIVE', 0.0)
        neg = scores.get('NEGATIVE', 0.0)
        sentiment = pos - neg

        if abs(pos - neg) < 0.15:
            label = "NEUTRAL"
        elif pos > neg:
            label = "SUPPORT"
        else:
            label = "CRITICIZE"

        pos_scores.append(pos)
        neg_scores.append(neg)
        sentiments.append(sentiment)
        labels.append(label)

    df['pos_score'] = pos_scores
    df['neg_score'] = neg_scores
    df['sentiment_score'] = sentiments
    df['label'] = labels

    # 5. Insert results into sentiment_score table
    insert_query = """
        INSERT INTO sentiment_score
        (username, comment, discussion_topic, comment_clean, pos_score, neg_score, sentiment_score, label)
        VALUES (%s,%s,%s,%s,%s,%s,%s,%s)
    """
    for _, row in df.iterrows():
        cursor.execute(insert_query, (
            row['username'],
            row['comment'],
            row['discussion_topic'],
            row['comment_clean'],
            row['pos_score'],
            row['neg_score'],
            row['sentiment_score'],
            row['label']
        ))

    conn.commit()
    cursor.close()
    conn.close()

    flash("Sentiment analysis completed and saved!", "success")
    return redirect(url_for("view_sentiment_score"))