sanitize_ml_labels

In this release, I have added support for properly sanitizing hyphenated words by integrating a lookup to a massive 45k vocabulary of hyphenated words in the English language.

Previously, hyphenated words written in the English language with hyphens were incorrectly stripped of their hyphens.

For instance, previously, non-existent-edges-in-graph would have become Non existent edges in graph. Now we handle the hyphenation correctly, and the result is Non-existent edges in graph.

In this release, we have moved the metrics that are normalized in a [-1, 1] range from the [0, 1] metrics (which of course was the wrong position) to the "absolutely normalized" set of metrics.

Such metrics include so far:

• Matthews correlation coefficient
• Markedness

In this new version of sanitize ML labels, I have:

• Added support for more terms to auto normalize
• Removed some metrics wrongly added to the "normalized" set

In this version, we introduce support for COVID19-related acronyms.