A Dopplr puzzle: the answer
Yesterday I asked what a particular list of English words had in common. Here’s the answer:
Mobile, AL, United States | Page, ND, United States | Nice, France | For, Norway | Name, Mozambique | Union, Philippines | Best, Netherlands | Dollar, United Kingdom | Tours, France | Split, Croatia | Accident, MD, United States | Mile, Tanzania | List, Germany | Central, SC, United States | Deal, NJ, United States | Deposit, NY, United States | Trip, Romania | This, France | Back, United Kingdom | Carry, Haiti | Center, MO, United States | Start, Russia | Kilo, Finland | Purchase, NY, United States | Transfer, PA, United States | Goès, France | Plan, Spain | Media, PA, United States | Sale, Australia | Downtown, PA, United States | Fully, Switzerland | Luck, WI, United States | Normal, IL, United States | Comfort, TX, United States | Call, MO, United States | Local, MO, United States
Every one of these words occurs often in emails sent by the travel industry. And every word is also the name of a place somewhere in the world. Congratulations to commenter Smyles yesterday for guessing most of the connection.
The reason these words are important to us is that they are the “stop” list we use to filter emails before trying to interpret trip information from them. Without this list, we’d see a phrase like “accident insurance” or “for your comfort” and offer you trips to the towns of Accident, For or Comfort.
I’m fascinated by how complicated the systems for recognising trips are. It had never occurred to me that words like this could cause problems.
Great series of posts.
[...] What do the following words have in common? mobile, page, nice, for, name, union, best, dollar, tours, split, accident, mile, list, central, deal, deposit, trip, this, back, carry, center, start, kilo, purchase, transfer, goes, plan, media, sale, downtown, fully, luck, normal, comfort, call, local addthis_url = ‘http%3A%2F%2Flaurent.pierssens.com%2F2008%2F08%2F15%2Fim-nowhere-near-ok-usa%2F’; addthis_title = ‘I%26%238217%3Bm+Nowhere+near+OK%2C+USA’; addthis_pub = ‘lfp’; [...]
I did come across Nowhere, OK, USA a few weeks ago but never thought there would be some many.
It sounds like you need to treat these words as not just “stop” words. (Where I work, we have similar problems with a set of “trouble” companies, e.g. Target, Google, Iceland…). For example, when you find a word like “nice” or “transfer”, you could look to the left and the right of it, to see if there are typical words associated with a place, to determine which sense of the word is being used in the email. It might also be helpful to see if the words are capitalized in the middle of sentence. And, if you have the staff, you could even send these emails for manual review.
Nice to see a window into Dopplr’s workings…
Unfortunately you forgot to mention Fucking in Austria, and Petting in Bavaria ;-)
[...] built up the top 1000 destinations, but also had to build a ’stop list’ of the common words in emails that are also town names, e.g. [...]