Scatter text interpretation

[albin02t]

Hi, First of the plot looks great and it is quite informative and intuitive.
But I'm having a hard time to interpret and understand some of the parameters as they aren't mentioned or touched on much in the README file and i get lost in the file.
So, if you don't mind can you explain:

1. How are the scores displayed on hover are calculated and what do they signify. I understand how the scaled f-scores are calculated each for the 2 classes.
2. What is the transform parameter used for and its values like st.Scalers.percentile_dense
3. term_scorer parameter and what does st.RankDifference() do
4. sort_by_distance
5. In the plot shown below, you can see that it is colored blue but score is a -0.015
   

[JasonKessler]

How are the scores displayed on hover are calculated and what do they signify. I understand how the scaled f-scores are calculated each for the 2 classes.

The scores, unless the term_scorer parameter to produce_scattertext_explorer is used, correspond to the dense rank difference.

What is the transform parameter used for and its values like st.Scalers.percentile_dense
term_scorer parameter and what does st.RankDifference() do sort_by_distance

The transform parameter defines how a terms's frequency is transformed into an axis position. The simplest approach, scale, places the minimum frequency at the origin and the maximum frequency at (1,0) or (0,1). The term_scorer determines the color of a point (dark blue at 1, and dark red at -1) as well as the score seen on hover.

sort_by_distance indicates whether the category-based term lists are based on the distance of each point to (0,1) or (1,0) (if True) or based on the scores computed by the term_scorer.

In the plot shown below, you can see that it is colored blue but score is a -0.015

That's the dense rank difference.

[albin02t]

Really appreciate you reaching back so quick, but can you clarify a few more things.

1. How is the dense rank difference calculated?
2. Is RankDifference() different from dense rank difference and how?
3. If the term_scorer scores the terms as dark blue as 1 and dark red -1, then in the plot above the word 'goes' is colored blue but the score is a -0.015. Could you explain how could that be?

Just starting out and thinking of implementing the scatter text in a product, so thanks for the help

[JasonKessler]

Please look at the source code and/or documentation for these questions.