I've been using and studying smlar -- any way I can help to extend this with it/vice versa?

[poppingtonic]

I updated smlar for postgresql 9.6, and this project seems very interesting given smlar's limitations:

• sometimes, the smlar(textarray,textarray) function in tfidf mode will return multiple results with the exact same rank, making it seem like there's no normalization. Does smlar do @@-style matching for each array element, or is there another element of similarity involved?
• It's difficult to get consistent results with smlar in tfidf mode. Maybe I don't understand smlar completely.
• I want to add a statistical autocorrect method, similar to this, that uses smlar.stattable and row-level triggers that update it based on changes in the fields that feed data to the table's tsvector field.

Thoughts?