April 3, 2009
We started nextstop.com with the idea that we could make it significantly easier and more fun to discover new and interesting things to do anywhere in the world, based on recommendations from people who know a place well. Whether it was a neat museum, a hidden local restaurant, or a great place to go shopping we wanted to make it super easy and fun for people to share recommendations for their favorite places, wherever they might live.
The trick of course was in how to do this. It was important for us to combine ease of making a recommendation -- our goal was that it should be as simple as entering the name of a place, and a few sentences about why you liked it -- with rich information about a place so it was really useful to others -- photos, contact information, maps, etc. The solution, not surprisingly since I'm writing here, was to use a number of Google's APIs to gather related information about the recommendation and make it easy for our members to include it in their recommendation.
You can best see how this works by going through our recommendation flow, or watching the video below.
Let me walk you through how this is working under the hood:
1) When the page loads, the first thing we do is use the Google loader to load the JQuery and JQuery UI libraries, as well as Google Maps. As part of this, we also grab the user's current location using google.loader.ClientLocation and store the lat/lng if available to use later.
2) In step 1, we ask the user for what's being recommended. We use this string to do a Google local search for business listings and KML results that match, using the user's current location to bound the local search by setting the sll and sspn parameters. Between local business listings and KML results, we can offer incredible global coverage of everything from restaurants to tourist attractions to hole-in-the-wall bars and clubs. We're using the JSON version of the local search API, which we call from our servers using Python's urlopen() so that we can supplement the results with our own database of results.
3) In step 2 we do an image search for related images using Google's image search API. While we let users change the search terms to find just the right picture, often our default image search (which combines the name of the place and a city name) returns great results. There are photos of almost everything, so you can even recommend a particular dish at a restaurant in Taipei and have the photos to go along with it.
4) In step 3, we ask for a few sentences about why that place or activity really stands out to them. After the recommendation has been submitted, we use the Google Language APIs to detect the language of the recommendation, which we can later use to filter content by your language, and we hope to someday integrate the ability to translate recommendations into your language of choice.
Its a very simple and fast process for the user making the recommendation, but the result is a recommendation with address, phone number, map, and photo that is really useful to another user looking to discover something new.
We've built our whole product around the Google APIs, and feel like we're just scratching the surface of what's possible. We're planning to let users add other information (like related websites, searches, news, etc.) using Google's APIs as well.
We'll be at Google I/O on May 27-28 talking about what we've done so far, and will hopefully have a few new uses of the Google APIs to show off at that time. Please come say hello -- we'd love to hear your feedback on nextstop, or share tips on using the Google APIs. You can also check out some of the places recommended near the Moscone center, or add a few of your own!
Carl Sjogreen (co-founder, nextstop.com)