Humaiz Akbar | OpinionAI

Overview

OpinionAI is an innovative AI-powered tool that generates dynamic conversational debates and analyzes the sentiment behind them. Built for CodeFest 25 in collaboration with Asjad, Hanzala Nadeem, and Uqba Nadeem, OpinionAI brings discussions to life while uncovering the emotions within the text.

AI Integration: We leveraged Claude AI through sophisticated prompt engineering techniques to generate balanced, nuanced debates on complex topics. Our approach focused on crafting precise prompts that would elicit comprehensive, multi-perspective responses while maintaining neutrality.

Features

Debate Generation

Input any topic, and OpinionAI crafts a lively, back-and-forth debate featuring diverse perspectives.

Sentiment Analysis

Analyzes the emotional tone of topics—whether positive, negative, or neutral—based on social media sentiment.

AI Integration

Leverages advanced natural language processing to create realistic, nuanced conversations on complex topics.

Social Media Insights

Provides insights into how topics are perceived across social media platforms.

Technology Stack

Python

Natural Language Processing

Machine Learning

Web Development

Implementation

OpinionAI combines two powerful features to create a one-of-a-kind experience. The backend is built with Python, utilizing advanced NLP techniques to generate debates and analyze sentiment.

Prompt Engineering

A critical aspect of our implementation was the development of sophisticated prompt engineering techniques for Claude AI. We designed a multi-stage prompting system that:

Establishes clear debate parameters to ensure balanced representation
Structures prompts to generate coherent, logical arguments from multiple perspectives
Implements context-aware follow-up prompts that maintain conversation continuity
Uses carefully crafted system prompts to maintain neutrality and factual accuracy

# Example of sentiment analysis implementation
import nltk
from nltk.sentiment import SentimentIntensityAnalyzer

def analyze_sentiment(text):
    sia = SentimentIntensityAnalyzer()
    sentiment_score = sia.polarity_scores(text)
    
    if sentiment_score['compound'] >= 0.05:
        return "Positive", sentiment_score
    elif sentiment_score['compound'] <= -0.05:
        return "Negative", sentiment_score
    else:
        return "Neutral", sentiment_score
        
# Example usage
topic = "Climate Change"
debate_text = generate_debate(topic)  # Function to generate debate
sentiment, scores = analyze_sentiment(debate_text)

print(f"The sentiment on {topic} is: {sentiment}")
print(f"Sentiment scores: {scores}")

The debate generation component uses Claude AI with our custom prompt engineering approach to create realistic conversations that present multiple viewpoints on a given topic.

# Example of prompt engineering for Claude AI
def create_debate_prompt(topic, perspective_a=None, perspective_b=None):
    """
    Create a structured prompt for Claude AI to generate debate content
    with balanced perspectives.
    """
    system_prompt = """You are a neutral debate moderator. Your role is to present 
    balanced arguments from multiple perspectives on controversial topics. 
    Ensure factual accuracy and fair representation of all viewpoints."""
    
    if perspective_a is None and perspective_b is None:
        # Initial prompt to start debate
        user_prompt = f"""Generate a balanced debate on the topic: {topic}
        
        Present two distinct perspectives (A and B) with thoughtful, nuanced arguments.
        Each perspective should:
        - Present clear reasoning
        - Use factual evidence where appropriate
        - Acknowledge complexity
        - Avoid strawman arguments
        
        Format as:
        Perspective A: [argument]
        Perspective B: [argument]"""
    else:
        # Follow-up prompt for continuing debate
        user_prompt = f"""Continue the debate on {topic}.
        
        Previous arguments:
        Perspective A: {perspective_a}
        Perspective B: {perspective_b}
        
        Generate the next exchange, with each perspective responding to the other's points.
        Maintain logical consistency with previous arguments while introducing new supporting points."""
    
    return system_prompt, user_prompt

Challenges & Solutions

Balanced Perspectives

Creating balanced, fair representations of different viewpoints was challenging. We implemented a perspective-neutral approach that focuses on presenting arguments without inherent bias, allowing users to form their own opinions.

Sentiment Accuracy

Accurately determining sentiment across diverse topics required sophisticated analysis. We combined multiple sentiment analysis techniques and trained our models on diverse datasets to improve accuracy across different domains and contexts.

Future Enhancements

Integration with real-time social media data for up-to-date sentiment analysis
Multi-language support for global topic analysis
Interactive debate feature allowing users to participate in AI-generated discussions
Topic recommendation system based on trending discussions

OpinionAI