6. Pipelines, Processes, Docs, and Words

Tip

See notebook https://github.com/cltk/cltk/blob/master/notebooks/CLTK%20data%20types.ipynb for a detailed walkthrough of CLTK data types.

The CLTK contains four important, native data types:

• cltk.core.data_types.Word: Contains all processed information for each word token. Has attributes including Word.string, Word.lemma, Word.pos, Word.governor, and Word.embedding. A Process adds data to each Word. See notebook https://github.com/cltk/cltk/blob/master/notebooks/CLTK%20Demonstration.ipynb for a full demonstration of what kind of information is stored in Word.

• cltk.core.data_types.Sentence: Contains sentence_embeddings a weighted average of the word embeddings of the sentence.

• cltk.core.data_types.Doc: Contains Doc.raw, which is the original input string to NLP().analyze(), and Doc.words, which is a list of Word objects. It is the input and output of each Process and final output of NLP(). See notebook https://github.com/cltk/cltk/blob/master/notebooks/CLTK%20Demonstration.ipynb for a full demonstration of what kind of information is stored in Doc

• cltk.core.data_types.Process: Takes and returns a Doc. Each process does some processing of information within the Doc, then annotates each Word object at Doc.words.

• cltk.core.data_types.Pipeline: Has a list of Process objects at Pipeline.processes. Predefined pipelines have been made for some languages (Languages), while custom pipelines may be created for these languages or other, different languages. See notebook https://github.com/cltk/cltk/blob/master/notebooks/Make%20custom%20Process%20and%20add%20to%20Pipeline.ipynb for an example creating a new Process and adding it to a custom Pipeline. For an illustration of how Process objects inherit from one another, see figure Inheritance of Pipeline class.

Inheritance of Pipeline class