The content of this blog is my personal opinion only. Although I am an employee - currently of Nvidia, in the past of other companies such as Iagination Technologies, MIPS, Intellectual Ventures, Intel, AMD, Motorola, and Gould - I reveal this only so that the reader may account for any possible bias I may have towards my employer's products. The statements I make here in no way represent my employer's position, nor am I authorized to speak on behalf of my employer. In fact, this posting may not even represent my personal opinion, since occasionally I play devil's advocate.

See http://docs.google.com/View?id=dcxddbtr_23cg5thdfj for photo credits.

Monday, October 01, 2012

Testing Mail Sorting Rules

I used to depend heavily, very heavily, on email sorting rule.  Back in the old days, when I used MH on UNIX. I can't even remember the mail filters programs I used.

As I was forced, against my will, to use Outlook, I used rules less.  Not because I wanted to - but because (1) Outlook's rules are much less reliable, and (2) the "development environment" for Outlook rules is much less good than text files on UNIX.  No version control.  No diff.

Basically, I tried using Outlook rules.  But they were so much hassle, so I gave up.

(Anecdote: at Intel I kept pestering IT for help with rules.  After I exhausted them, they told me about this power user they had heard about, who knew all about Outlook rules.  IT person #1 did not know who it was, so sent me to IT person #2, who sent m to ... Eventually we found this power user.  It was me, myself, and I. As incarnated before I left Intel for AMD. My reincarnation after I returned from AMD to Intel was referred to my original incarnation.)

GMail was better for a while.  The rules are slightly better than Outlook's.  But, mainly, the Bayesian importance filter was able to help a lot.

But the email onslaught grows.  So I am back thinking about emal rules again.  In both Outlook and Gmail.


Minimum requirement: version control.  It is necessary to track what works, and what doesn't.  Comments.


Thinking about why email rules are such a hassle:


Outlook rules in particular are a pain, because it is had to create a test input set.   The rules that move items to a particular folder, well, move them.   Thus destroying the test input set.

I suppose the rules could be tested in a special account, and then copied.  But that has hassles.

If the rules could be made conditional - only optionally moving....

 Having the rules apply labels or categories rather than move helps.  

Gmail's folders, of course, are really just labels or categories.


What I really want is multiway Bayesian classification: not just important/unimportant, but applying tags other than important.  Machine learning to learn what labels and categories I want to apply.


For now, though I am stuck with Outlook.


No comments: