nlppln 0.3.0

nlppln is a python package for creating NLP pipelines using Common Workflow Language (CWL). It provides steps for (generic) NLP functionality, such as tokenization, lemmatization, and part of speech tagging, and helps users to construct workflows from these steps.

A text processing step consist of a (Python) command line tool and a CWL specification to use this tool. Most tools provided by nppln wrap existing NLP functionality. The command line tools are made with Click, a Python package for creating command line interfaces.

To create a workflow, you have to write a Python script:

from nlppln import WorkflowGenerator

with WorkflowGenerator() as wf:
  txt_dir = wf.add_inputs(txt_dir='Directory')

  frogout = wf.frog_dir(in_dir=txt_dir)
  saf = wf.frog_to_saf(in_files=frogout)
  ner_stats = wf.save_ner_data(in_files=saf)
  new_saf = wf.replace_ner(metadata=ner_stats, in_files=saf)
  txt = wf.saf_to_txt(in_files=new_saf)

  wf.add_outputs(ner_stats=ner_stats, txt=txt)

  wf.save('anonymize.cwl')

The resulting workflow can be run using a CWL runner, such as cwltool:

cwltool anonymize.cwl --txt_dir /path/to/directory/with/txt/files/

For creating new (e.g., project specific) NLP functionality, you can use nlppln-gen to generate boilerplate (i.e., empty) command line tools and CWL specifications.

The full documentation can be found on Read the Docs.

pip install nlppln