The Pimlico Processing Toolkit is a toolkit for building pipelines of tasks for processing large datasets (corpora). It is especially focussed on processing linguistic corpora and provides wrappers around many existing, widely used NLP (Natural Language Processing) tools.
Pimlico is written in Python and can be run using Python >=2.7 or >=3.6. This means you can write your own processing modules using either Python 2 or 3.
These are the docs for the release candidate for v1.0.
This brings with it a big project
to change how datatypes work internally (previously in branch
and requires all datatypes and modules
to be updated to the new system.
Modules marked with
!! in the docs are waiting to be updated and
don’t work. Other known outstanding tasks are marked with todos:
full todo list.
These issues will be resolved before v1.0 is released.
It makes it easy to write large, potentially complex pipelines with the following key goals:
- to provide clear documentation of what has been done;
- to make it easy to incorporate standard NLP tasks,
- and to extend the code with non-standard tasks, specific to a pipeline;
- to support simple distribution of code for reproduction, for example, on other datasets.
The toolkit takes care of managing data between the steps of a pipeline and checking that everything’s executed in the right order.
The core toolkit is written in Python. Pimlico is open source, released under the GPLv3 license. It is available from its Github repository. To get started with a Pimlico project, follow the getting-started guide.
Pimlico is short for PIpelined Modular LInguistic COrpus processing.
More NLP tools will gradually be added. See my wishlist for current plans.