Introduction
You’ll learn how to build a custom NLP project with Vertex AI Workbench by using TensorFlow.
Unlike AutoML
, which is a no-code solution, with custom training you must code on your own.
Luckily, TensorFlow
provides many libraries and pre-built functionalities that you can rely on to make the coding easier.
You’ll start with tokenization, which is a technique used to break down a text into smaller pieces, so that a computer can digest them.
You’ll then be introduced to one-hot encoding and bag-of-words, the two basic but fundamental techniques used to turn words into numbers or vectors
After that, you’ll proceed to more advanced techniques such as word embeddings
and its most important model, word2vec
.
Finally, you’ll advance to transfer learning and reusable embeddings,
followed by a hands-on lab where you’ll use reusable embeddings.