Apply Data Augmentation

[Task of Domain Adoption and Customization]

Purpose

Generate more training examples to train a better performing model

Description

Data Augmentation describes a method to generate new trainings examples with different strategies: e.g., introducing noise to data points (typo, removing words). Depending on the use case different strategies are appropriate. The goal is to train a better performing and robust model. This tasks is marked as technical task. Thus, the definition of Data Augmentation Strategies with corresponding implementations triggers the data augmentation step as a part of the training process. If the augmentation process is too computation-intensive, the data augmentation should be done beforehand and versioned as new corpus.

Steps

• None

Artifacts

Requires

• Data Augmentation Strategies

Creates

• Augmented Training Dataset

• Evaluation Dataset

• Test Dataset

Responsible Role

• None