Text Classification Call

Text Classification

Text Classification calls are calls that can take text and give weights of confidence back of how well this text fits a classification. Famous providers are Hugging Face models like nlptown/bert-base-multilingual-uncased-sentiment for sentiment analysis or protectai/deberta-v3-base-prompt-injection for prompt injection detection.

Example normalized Text Classification call

The following is an example of sending the text "This product is absolutely wonderful!" into Hugging Face using the nlptown/bert-base-multilingual-uncased-sentiment model and getting the classification back in form of an array of TextClassificationItem.

use Drupal\ai\OperationType\TextClassification\TextClassificationInput;

$input = new TextClassificationInput('This product is absolutely wonderful!');
/** @var array $classification_items */
$classification_items = \Drupal::service('ai.provider')->createInstance('huggingface')->textClassification($input, 'nlptown/bert-base-multilingual-uncased-sentiment', ['my-custom-call'])->getNormalized();

foreach ($classification_items as $item) {
  // If the text is classified as very negative with high confidence, flag it.
  if ($item->getLabel() == '1 star' && $item->getConfidenceScore() > 0.7) {
    throw new \Exception('This text has been classified as very negative.');
  }
}

Text Classification Interfaces & Models

The following files defines the methods available when doing a text classification call as well as the input and output.

• TextClassificationInterface.php
• TextClassificationInput.php
• TextClassificationOutput.php

Text Classification Explorer

If you install the AI API Explorer, you can go configuration > AI > AI API Explorer > Text Classification Explorer under /admin/config/ai/explorers/ai-text-classification to test out different calls and see the code that you need for it.