Pre-built APIs lets you leverage the NLP problems that were solved and the NLP models that were trained by Google, so you don’t need to build your own NLP models from scratch if you lack training data and machine learning expertise inside of the company.

AutoML is a no-code solution, so you can build your own NLP models on Vertex AI through a point-and-click interface.

Through custom training, you can code your very own NLP environment to train and deploy an NLP model with Vertex AI Workbench, a notebook tool.

This approach provides you with flexibility and control over the entire process.

Training data size:

Pre-built APIs do not require any training data, AutoML requires small to medium size training data, and custom training normally requires a large amount of data.

Machine learning and coding expertise:

Pre-Built APIs and AutoML are user-friendly with low requirements, but Custom training has a high requirement of ML and coding expertise.

Flexibility to tune hyperparameters:

At the moment, you can’t tune hyperparameters with Pre-built APIs or AutoML; however, you can experiment with hyperparameters with custom training.

The time to train a model:

Pre-built APIs require no time to train a model, because they directly use pre-built models from Google.

For the other two options, both take time to train and the training time depends on the projects and the resources.

Selecting the best option depends on your business needs and the expertise in NLP and ML.

If you have little ML/NLP experience and no training data, using pre-built APIs is likely the best choice.

Pre-built APIs address common perceptual tasks in natural language processing such as translation and document analysis.

They are ready to use without any NLP and ML expertise or model development effort.

If you want to build custom models with your own training data and spend minimal time coding, then AutoML is your choice.

AutoML provides a no-code solution to let you focus on business problems instead of the underlying model architecture and ML provisioning.

If you want full control of the NLP workflow, for example, code your own model development environment and experiment with hyperparameters, Vertex AI custom training lets you train and serve custom models with code on Vertex AI Workbench.