Redaguoti

Bendrinti naudojant


Evaluation metrics for orchestration workflow models

Your dataset is split into two parts: a set for training, and a set for testing. The training set is used to train the model, while the testing set is used as a test for model after training to calculate the model performance and evaluation. The testing set isn't introduced to the model through the training process, to make sure that the model is tested on new data.

Model evaluation is triggered automatically after training is completed successfully. The evaluation process starts by using the trained model to predict user defined intents for utterances in the test set, and compares them with the provided tags (which establishes a baseline of truth). The results are returned so you can review the model’s performance. For evaluation, orchestration workflow uses the following metrics:

  • Precision: Measures how precise/accurate your model is. It's the ratio between the correctly identified positives (true positives) and all identified positives. The precision metric reveals how many of the predicted classes are correctly labeled.

    Precision = #True_Positive / (#True_Positive + #False_Positive)

  • Recall: Measures the model's ability to predict actual positive classes. It's the ratio between the predicted true positives and what was actually tagged. The recall metric reveals how many of the predicted classes are correct.

    Recall = #True_Positive / (#True_Positive + #False_Negatives)

  • F1 score: The F1 score is a function of Precision and Recall. It's needed when you seek a balance between Precision and Recall.

    F1 Score = 2 * Precision * Recall / (Precision + Recall)

Precision, recall, and F1 score are calculated for:

  • Each intent separately (intent-level evaluation)
  • For the model collectively (model-level evaluation).

The definitions of precision, recall, and evaluation are the same for intent-level and model-level evaluations. However, the counts for True Positives, False Positives, and False Negatives can differ. For example, consider the following text.

Example

  • Make a response with thank you very much
  • Call my friend
  • Hello
  • Good morning

These are the intents used: CLUEmail and Greeting

The model could make the following predictions:

Utterance Predicted intent Actual intent
Make a response with thank you very much CLUEmail CLUEmail
Call my friend Greeting CLUEmail
Hello CLUEmail Greeting
Goodmorning Greeting Greeting

Intent level evaluation for CLUEmail intent

Key Count Explanation
True Positive 1 Utterance 1 was correctly predicted as CLUEmail.
False Positive 1 Utterance 3 was mistakenly predicted as CLUEmail.
False Negative 1 Utterance 2 was mistakenly predicted as Greeting.

Precision = #True_Positive / (#True_Positive + #False_Positive) = 1 / (1 + 1) = 0.5

Recall = #True_Positive / (#True_Positive + #False_Negatives) = 1 / (1 + 1) = 0.5

F1 Score = 2 * Precision * Recall / (Precision + Recall) = (2 * 0.5 * 0.5) / (0.5 + 0.5) = 0.5

Intent level evaluation for Greeting intent

Key Count Explanation
True Positive 1 Utterance 4 was correctly predicted as Greeting.
False Positive 1 Utterance 2 was mistakenly predicted as Greeting.
False Negative 1 Utterance 3 was mistakenly predicted as CLUEmail.

Precision = #True_Positive / (#True_Positive + #False_Positive) = 1 / (1 + 1) = 0.5

Recall = #True_Positive / (#True_Positive + #False_Negatives) = 1 / (1 + 1) = 0.5

F1 Score = 2 * Precision * Recall / (Precision + Recall) = (2 * 0.5 * 0.5) / (0.5 + 0.5) = 0.5

Model-level evaluation for the collective model

Key Count Explanation
True Positive 2 Sum of TP for all intents
False Positive 2 Sum of FP for all intents
False Negative 2 Sum of FN for all intents

Precision = #True_Positive / (#True_Positive + #False_Positive) = 2 / (2 + 2) = 0.5

Recall = #True_Positive / (#True_Positive + #False_Negatives) = 2 / (2 + 2) = 0.5

F1 Score = 2 * Precision * Recall / (Precision + Recall) = (2 * 0.5 * 0.5) / (0.5 + 0.5) = 0.5

Confusion matrix

A Confusion matrix is an N x N matrix used for model performance evaluation, where N is the number of intents. The matrix compares the actual tags with the tags predicted by the model. This gives a holistic view of how well the model is performing and what kinds of errors it is making.

You can use the Confusion matrix to identify intents that are too close to each other and often get mistaken (ambiguity). In this case consider merging these intents together. If that isn't possible, consider adding more tagged examples of both intents to help the model differentiate between them.

You can calculate the model-level evaluation metrics from the confusion matrix:

  • The true positive of the model is the sum of true Positives for all intents.
  • The false positive of the model is the sum of false positives for all intents.
  • The false Negative of the model is the sum of false negatives for all intents.

Next steps

Train a model in Language Studio