You are working on a prototype of a text classification model in a managed Vertex AI Workbench notebook. You want to quickly experiment with tokenizing text by using a Natural Language Toolkit…

Installing from a terminal is possible but less integrated for quick experimentation within the notebook's active kernel, and the `--user` flag is important for user-specific installations.

Writing a custom Dataflow job is a solution for large-scale data processing, not for quickly adding a library to a local notebook environment for prototyping.

Creating a new custom image for a notebook is a more involved process suitable for complex, persistent dependencies or production environments, not for rapid, ad-hoc experimentation.

Running `!pip install nltk --user` directly within a Jupyter cell in a Vertex AI Workbench notebook is the quickest and most convenient way to install a Python library for immediate use and experimentation without requiring environment restarts or custom image creation.