Text Classification Template

Table of Contents

Overview

Complexity: Simple (2 blocks) Use Case: Classify text into predefined categories with confidence scores

This template automatically categorizes text content into one of five predefined categories: environment, technology, health, finance, or sports. Each classification includes a confidence score.

Pipeline Architecture

┌──────────────────┐     ┌──────────────────┐
│   Structured     │ ──► │       JSON       │
│    Generator     │     │    Validator     │
└──────────────────┘     └──────────────────┘

Input: content
  ↓
+ generated (category, confidence)
  ↓
+ valid, parsed_json

Blocks:

  1. StructuredGenerator - Classifies text with schema enforcement
  2. JSONValidatorBlock - Validates category enum and confidence range

Seed Format

Use the simplified content field in metadata:

{
  "repetitions": 2,
  "metadata": {
    "content": "Solar panels convert sunlight into electricity for homes and businesses."
  }
}

Multiple seeds example:

[
  {
    "repetitions": 1,
    "metadata": {
      "content": "The stock market reached record highs today as investors responded positively."
    }
  },
  {
    "repetitions": 1,
    "metadata": {
      "content": "New breakthrough in cancer treatment shows promising results in clinical trials."
    }
  }
]

Output Format

Schema:

{
  "category": "enum [environment, technology, health, finance, sports]",
  "confidence": "number (0-1)"
}

Example outputs:

{
  "category": "environment",
  "confidence": 0.95
}
{
  "category": "health",
  "confidence": 0.88
}

Default Categories

The template includes 5 predefined categories:

  1. environment - Climate, sustainability, nature, energy
  2. technology - Software, hardware, AI, innovation
  3. health - Medicine, fitness, wellness, treatment
  4. finance - Markets, economy, banking, investment
  5. sports - Athletics, competitions, teams, games

Use Cases

Perfect for:

  • Content categorization for blogs or news sites
  • Automated tagging systems
  • Dataset labeling for training models
  • Organizing unstructured text collections

Not ideal for:

  • Binary yes/no classification (customize categories)
  • Sentiment analysis (customize to positive/negative/neutral)
  • Multi-label classification (requires custom block)

Customization

Modify the template in lib/templates/text_classification.yaml:

Change categories (e.g., for sentiment analysis):

json_schema:
  properties:
    category:
      type: string
      enum: ["positive", "negative", "neutral"]
    confidence:
      type: number
      minimum: 0
      maximum: 1

Add more categories:

enum: ["environment", "technology", "health", "finance", "sports", "education", "entertainment"]

Adjust temperature for more/less deterministic results:

temperature: 0.1  # Very deterministic
temperature: 0.7  # More creative/varied

Customize the prompt:

user_prompt: "Classify the sentiment of this customer review as positive, negative, or neutral. Review: {{ content }}"

Common Use Case Examples

Sentiment Analysis:

# Categories: positive, negative, neutral
# Use for: Customer reviews, feedback, social media

Support Ticket Routing:

# Categories: billing, technical, account, general
# Use for: Automated ticket classification

Content Moderation:

# Categories: safe, questionable, inappropriate
# Use for: User-generated content filtering