تخصيص المسار

A spaCy pipeline is just an ordered list of components. Text flows through each one, gaining tags, parses, and entities along the way.

Check the active steps with nlp.pipe_names. It lists every component in order, so you always know what is running.

Skip work you do not need with disable. Turning off the parser when you only want entities makes processing noticeably faster.

For a quick run, use select_pipes as a context manager. It switches components off, then restores them when the block ends.

You extend the pipeline with add_pipe. spaCy ships ready-made components like sentencizer that you can drop right in.

Position matters. With first or before, you place a component exactly where it needs to run in the sequence.

You can register your own logic as a component. A simple function takes a Doc, tweaks it, and returns it for the next step.

Once registered, add your component by name. From then on it runs automatically every time you call nlp().

Changed your mind? remove_pipe takes a component out cleanly, leaving the rest of the pipeline untouched.

You can shuffle steps later too. Combining remove and add lets you reorder the pipeline until the flow is exactly right.

Customizing the pipeline keeps spaCy lean and tailored. You run only what your task needs, plus any logic unique to your domain.

How do you add a component to a spaCy pipeline?

You inspect steps with pipe_names, drop work with disable, and shape the flow using add_pipe and remove_pipe. The pipeline bends to fit your task. 🛠️