AWS Comprehend simplifies custom model retraining and management

Amazon Comprehend announced self-service flywheel APIs to simplify the retraining and version management of custom Comprehend models. [Amazon Comprehend](/comprehend/) is a Natural Language Processing (NLP) service that provides pre-trained and custom APIs to derive insights from textual data. Customers can bring their own data, and train custom Comprehend models to classify documents, and extract entities of interest for their specific business needs. Until today, customers needed manual processes for merging existing and new datasets to train new models, evaluating improvements in model performance, and managing model versions over time. Starting today, customers using the flywheel feature just need to provide the new dataset for retraining. The feature retrains the model by automatically merging existing and new datasets, displays the performance of the model against previous versions that it maintains, and enables customers to select the best version as the production model. Self-service flywheel is available for custom Comprehend Named Entity Recognition (NER) and Document Classification models that are trained using plain-text documents. Customers can create a Flywheel using CLI, API or the AWS Console. To learn more and get started, refer to the feature [overview blog](https://aws.amazon.com/blogs/machine-learning/introducing-the-amazon-comprehend-flywheel-for-mlops/) or visit the Amazon Comprehend [documentation](https://docs.aws.amazon.com/comprehend/latest/dg/flywheels.html) and [product page](/comprehend/).