A public web application collects large amounts of free-text user feedback. Product managers have labeled a large historical dataset into a fixed set of categories (for example, UI issues, performance issues, feature requests, chat issues). An ML engineer must automate multi-class classification of new feedback using Amazon SageMaker. Which SageMaker algorithm is the most appropriate for this task?

BlazingText is the most appropriate choice because it is a highly efficient algorithm for text classification tasks, including multi-class classification, and works well with large datasets. It supports both Word2Vec and FastText modes, making it versatile for generating word embeddings and performing classification. LDA and NTM are topic modeling algorithms, designed to discover abstract "topics" within a collection of documents, not to classify documents into predefined categories. While they can be part of a larger text processing pipeline, they are not direct classification algorithms. CatBoost is a gradient boosting decision tree algorithm primarily used for tabular data, not optimized for free-text classification.