Alright: without going full Bayesian statistics, it's probably best to make this generic:

1) Ruleset. A Rules collection contains a list of Rule. Rules(Of Rule)
2) A Rule contains the regular expression terms to apply. Integers can be applied to a tweet's overall score based on the success/failure of the Regex match.
3) This adds a lot more UI complexity and probably shouldn't be implemented, but maybe Rules can be linked; eg. if Rule A matches AND Rule B matches, then apply a modifier.

This is going to take a lot of editors to implement correctly. No Interfaces or anything like that - I don't want to go the FileSieve route as the TD userbase is vastly smaller - but just a collection of Panels. While the user won't know a difference, it won't be modular or extensible via plugins like FileSieve is.

This is a huge amount of work, so no other features will be added or worked upon until this is done; the tracker will look like activity has ceased on TD - but this isn't the case.

With no optimisations (eg. don't keep creating Regex objects; tweet.Count * rule.Count), it takes 49ms to run two rules on 21,000 tweets.

Quick optimisation later and now it takes 13ms.

A quick WIP look at the RuleSet class diagram. The Rule class is going to be a lot bigger.

RuleSet-ClassDiagram1.png (18,482 bytes)   

RuleSet-ClassDiagram1.png (18,482 bytes)   

Importing and exporting is in.

Tweet Scoring is in, albeit not entirely complete. Not all comparison fields are available but will be added in an update; see 0000490.

Exporting a Rule Set to the clipboard isn't yet implemented, either; see 0000489.