Update keyterms module

[bdewilde]

Description

• Move keyterm extraction functionality from a top-level keyterms module into a ke sub-package, and refactor+standardize its contents
  • all methods have similar args/options, and share code for selecting candidates, normalizing terms to strings, filtering to just the top-N key terms, and building term graphs
• Add new unsupervised keyterm extraction algorithms
  • YAKE: statistical method, implemented in ke.yake()
  • sCAKE: graph-based method, implemented in ke.scake()
  • PositionRank: graph-based method, implemented in ke.textrank() with parameter values given in the docstring
• Add new functionality for selecting candidate keyterms (in addition to n-grams method)
  • longest matching subsequence candidates: implemented in ke.utils.get_longest_subsequence_candidates()
  • pattern-matching candidates: implemented in ke.utils.get_pattern_matching_candidates()
• Significantly improve speed of SGRank and generally optimize all of these algorithms

Motivation and Context

Still hunting for the "perfect" unsupervised keyterm extraction algorithm, although all of these methods have pros/cons. A lit review of recent results pointed me towards YAKE and sCAKE.

How Has This Been Tested?

Added lots of tests, and they all pass.

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)

Checklist:

• My code follows the code style of this project.
• My change requires a change to the documentation, and I have updated it accordingly.