Dopplr import fail

I use Dopplr and TripIt, to keep track of traveling colleagues, update people on my own travel, and just generally simplify life. (I’ve blogged about each many times as well: see posts containing dopplr or TripIt).

I’m getting tired of Dopplr’s consistent FAIL on a common (for me) use case- a one day trip to NY on Delta.

Whenever I forward such a confirmation here’s what Dopplr does:

Thanks for sending us a message by email.

We automatically created a trip to Atlanta, GA, United States (from Newburyport, MA, United States) between February 10th and February 13th

We won’t share coincidences from this newly-created trip with your fellow travellers until February 18th. This is to give you a chance to check and correct any problems in interpretation.

If you’d like to check, go to [link removed]

Yours sincerely,
The Dopplr Team.

The problem is, I’m not going to or from Atlanta – that’s where Delta airlines headquarters is, sure, but it’s a long detour on the Boston->New York route.

I’m also not travelling Feb 10 to Feb 13, I’m leaving and returning on the 13th, and the trip was booked on the 10th.

I can (and do) go in and manually fix the trip in Dopplr, but in this type of case TripIt’s import feature just works.

Dopplr gets Email, Twitter, SMS import

One of the more popular posts on this blog is the one which describes how to import trips from TripIt into Dopplr, in order to avoid the re-entry tax. After all, as I wrote in my comparison of the two services last October, TripIt’s email import was the critical factor in my decision of how to manage this information:

Tripit’s mechanism for adding trips is superior. The ability to simply forward (or even set an automatic rule to forward) confirmation emails is a major step forward . . . Where TripIt seems better at pulling data in, Dopplr seems to be better so far at pushing their data out, or letting people pull it into other contexts.

Well, now Dopplr’s gone and added some new import mechanisms of their own. This post from the Dopplr blog (ok, it was posted back on July 8th, but it has been sitting in my queue to write about) lays out three new options: Twitter, SMS, and Email:

Today I’m really happy to say we’re taking the wraps off a number of new ways to get your future into Dopplr and share your travel information with those you trust: Dopplr by Twitter, SMS and… Email!

Although I love twitter as a notification service (a way of letting me know something relevant happened) I don’t see myself using it as a data input service. For those of you who would like to, just follow the dopplr user and send direct messages with your trips, like: d dopplr a trip to London July 28th to August 3rd. (Nicely, it also happily accepts @dopplr posts, in case you want to announce your trips as well as put them in dopplr). SMS is another option – you associate your SMS number with your Dopplr account and you can text message the same types of messages to Dopplr’s number.

Finally, they’ve got email working at trips@dopplr.com (wonder how many people will confuse plans@tripit.com with trips@dopplr.com – did they make plans@dopplr.com an alias?).

Interestingly, you can use the same kind of shorthand messages used for Twitter or SMS – “a trip to London July 28th to August 3rd” – or you can forward confirmation messages from booking services (which is how TripIt handles import). This is because Dopplr did not set out to parse all the complex formats used by different agencies, but took a simper approach, as explained by MattB:

There are an awful lot of ways to format a travel itinerary. When people asked us to extract trips from emails, we looked at our long history of e-tickets, confirmations and reservations, and scratched our heads.

Inspiration came in the shape of Apple’s last OS X release, Leopard, and an intriguing feature called “Data detectors“.

We realised that instead of creating a piece of code to decode every email format out there, we could look for patterns of dates and place names in the text (and later, other information too) and turn those into trips.

A happy side-effect of this approach is that as well as extracting information from automatic reservation emails, it works well with short text strings like “I’ll be in San Francisco from 3rd July to 7th July”. This means we can work with many hand-written emails, with Twitters, and with SMSes too.

Of course it won’t work with every variation under the sun (for example, it’s most reliable when an email contains just a return trip in a single hop), but we’ve had very satisfying results in our testing. And of course every email you send us will be added to our test suite so that our engine can get better and better over time.

In other words, rather than specifically targeting all the different potential formats, and parsing them in some structured way, Dopplr looks for some specific patterns in the text and tries to understand their meaning without knowing the format of the email in advance.

I wonder how different this is from what TripIt actually does behind the scenes – how much they plan for specific formats they know in advance – and how successful it will be “in the field.” For now it is enough to convince me to turn off my automated importing and give trips@dopplr.com a try on my next few confirm messages. Then, I can automate a rule in my email such that travel confirmations get auto-forwarded to both plans@tripit.com and trips@dopplr.com, and be sharing my travel plans painlessly.