Zephyr is taking feature requests or bug reports here. If you are looking to add a new pattern to watch for or room to be monitored, you will care about this file.
Actually @Benjamin, the current established fields of ML aren't very mature yet, I'm slightly glad we crafted out own logic than following known technologies.
Although you want to do actual text processing, if I may suggest - look into bag of words.
That is - just put all the words in a giant bag, and then use the TF-IDF metric and learn it with SVM, it's really simple and I'm sure there is a C# library for half of it.
I wrote a spam classifier in Python the other year for an ML course which was pretty simple - I can dig out the code if you're interested.
@BenjaminGruenbaum The problem with that is it'll easily miss spam with gibberish titles (e.g., "What better yesterday he", "Fun down said banana", etc.).
@Sam remember, the easiest thing everyone is doing is SVM with bag of words. If you don't know what a vector space is or what the kernel trick is - you better use a library.
> A perfect predictor would be described as 100% sensitive (i.e. all sick are identified as sick) and 100% specific (i.e. all healthy are identified as healthy)
