Swaroop C H

blog books about contact subscribe

Ideas are Cheap - Better email

26 Sep 2008

I was just browsing through this paper on "Generating summary keywords for emails using topics" and was fascinated by its usefulness, especially because it "selects words that describe each message in the context of existing topics rather than simply selecting keywords based on a single message in isolation."

Imagine if we had a smart email client which will automatically show the summary of the email in a few words rather than simply junk/not junk classification. That would help us a lot to triage our email.

On another tangent, how about an email client that sorts your inbox by importance and not by time? Importance can be automatically determined by how often you reply to the sender, what topics you reply fastest to, whether the sender is from the same company, etc.

We have websites like Amazon and Yahoo! that automatically customize their websites based on our usage patterns, why can't email clients do the same?

Email clients have been around for so long, can they become smarter than a sorted grid with folders?

P.S. While email is the big fish, there is a lot that can be done for personal information management.

Comments

Arvind says:

I think www.xobni.com is doing a great job in that space.

t3rmin4t0r says:

Determinism is something I very much value in machines. In fact, that's what's different about them from people in the first place. Fuzzy stuff like this, as useful as it might seem is sometimes completely annoying when it does not work.

I guess, I'm a glass half-full person. But getting it half-right is not very hard, it's just that half-right is not a milestone to evolve further from - except from inside a lab, before it is "born" into the real world. All or nothing problems require quantum leaps.

Devdas Bhagat says:

Ideas aren't cheap. Incremental ideas are cheap, quantum jumps are expensive.

If it was that simple for a machine to deal with messages in topic context, it would be nice.

But then, the challenge is to deal with those of us who have firehoses of email flowing in on various topics between the same set of people.

You would also need to be aware of different relationships between people, and the state of importance at that instance of time.

Of course, given that most companies still can't get threading right, sorting by importance is hard to ask for in a deterministic fashion.

Swaroop says:

@Arvind I've generally seen the description of xobni w.r.t. the people-aspect of things and not the organization aspect of things, but it is a step in this direction :)

@Gopal As long as they don't turn out to be powerful AI which in turn will build a Matrix, I'm okay with the feature being there and having a small checkbox saying turn it off :)

@Devdas Yes, it is a hard problem, but imagining a search index for most of the world wide web which will return results in seconds was a hard problem decades ago but is a reality today.

Feedback

There's no comment box, but please do email me or tweet me your thoughts and criticisms, and I will publish the relevant ones here.