After a model produces an output, there is a need to evaluate its quality quantitatively rather than by impression. Evaluation metrics give a number that can be compared across runs.

Evaluation metrics assist us in 3 tasks:

  1. Objective Comparisons: a common framework for assessing and comparing the performance of different models
  2. Model Selection: informed choices when picking the best performing model among several candidates
  3. Hyperparameter Tuning: assessing the impact of hyperparameter choices, so the best set can be selected

Types of Models

Text models are generally split into 2 types, and each type requires different kinds of evaluation metrices.

TypeWhat it doesExamples
Predictivetrained to make predictions or classifications from input dataclassification models, regression models
Generativetrained to create new data, usually textAI chatbots, translators, text summarisers

Metrics for predictive models:

  1. Confusion Matrix
  2. F1 Score
  3. Area Under Curve (AUC-ROC)

Metrics for generative models:

  1. Bilingual Evaluation Understudy (BLEU)
  2. Recall-Oriented Understudy for Gisting Evaluation (ROUGE)
  3. Metric for Evaluation of Translation with Explicit Ordering (METEOR)

Precision, recall and F1 can each be reported in 2 ways when there is more than 1 class, which is covered in Micro and Macro Metrics.

The outputs being measured here come from the models in Text Classification.