Final part of series on leveraging AI for emotional intelligence

In our previous posts, we explored how to train a custom AI model for recognizing emotions in text. Now, we're taking the next step: building a user-friendly web application that turns our powerful backend model into a tool anyone can use—no coding required.

From Code to Clarity: Making AI Accessible

While AI models are incredibly powerful, they often remain locked behind complex APIs and code. Our goal is to bridge this gap with a Streamlit application that provides an intuitive interface for our emotion classification model.

Application Screenshot Placeholder

This application allows users to:

  • Input text directly or upload files containing multiple entries
  • Process up to 50 text items simultaneously
  • View immediate classification results and track historical analyses
  • Export results for further processing or reporting

How It Works: The User Journey

Let's walk through the typical user experience:

  1. Setup: Users provide their Cohere API key and optionally specify a custom model ID (in our case, our fine-tuned emotion classifier)
  2. Input: Users can type directly into the chat interface or upload files (TXT, CSV, Excel)
  3. Classification: With a single click, the application sends the text to our model and retrieves emotional insights
  4. Results: The app displays both individual classification details and summary statistics
  5. History: All results are saved with timestamps for future reference

The Technical Magic Behind the Scenes

Our application is built on Streamlit, a Python framework that makes it easy to create data applications. The core functionality revolves around:

Flexible Input Processing 

def process_file_content(file_input):
    """Extract text content from uploaded file."""
    # Code determines file type and extracts text accordingly
    if file_name.endswith('.txt'):
        # Process text files
    elif file_name.endswith('.csv'):
        # Process CSV files
    # etc.

This function handles various file formats intelligently, extracting text that can be fed to our emotion classifier.

Seamless AI Integration

The application connects to Cohere's API with minimal code:

response = co.classify(
    inputs=string_inputs,
    model=model_id
)

This single function call sends our text to the fine-tuned model and returns structured classifications.

Intelligent Results Management

Results are organized into both immediate displays and historical records:

# Create results dataframe with full original text and timestamp
batch_results = [
    {
        'text': item.input,
        'prediction': item.prediction,
        'confidence': round(item.confidence, 4),
        'timestamp': current_time
    }
    for item in response.classifications
]

From Raw Text to Emotional Intelligence

What makes this application powerful is how it transforms simple text into emotional insights. For example, a customer service team might upload hundreds of support tickets and quickly identify:

  • Tickets expressing frustration that need immediate attention
  • Overall emotional trends in customer communications
  • Specific product features that trigger negative emotions
  • The emotional impact of recent company changes or announcements

Building Your Own Text Classification Interface

Want to create a similar application? Here's what you'll need:

  1. A Cohere API key - Sign up at cohere.com
  2. A fine-tuned model - Either train your own (see our previous post) or use Cohere's pre-trained options
  3. Basic Python knowledge - To understand and modify the Streamlit code
  4. Streamlit - Install with pip install streamlit

The full code for this application is available in our GitHub repository.

While we've focused on emotion detection, this same interface can be adapted for:

  • Content moderation (identifying inappropriate text)
  • Customer intent classification (purchase interest, support needs, etc.)
  • Document categorization (automatically sorting documents by type)
  • Language detection (identifying which language is being used)

Conclusion: Democratizing AI Access

The true power of AI isn't just in its capabilities but in its accessibility. By creating user-friendly interfaces like this Streamlit application, we make sophisticated emotional intelligence available to everyone in an organization—not just the data science team.

Build something cool today! 🚀🚀🚀