View the Full Mockup Sequence

Search Party Feature

Search Party is a web browser extension which "socializes" web searches, connecting people with ephemeral interests in particular topics together. What people do with that connection is up to them, but I imagine both strictly functional uses (sharing links to the most pertinent web pages, disseminating non-web information related to the topic, etc) and more sociable, entertainment oriented conversations (say, arguing about the relative merits of a Corvette vs. a BMW M3).

Existing chat systems are typically geared toward long-term communities comprised of people with a persistent (strong) interest in a topic. For example. the people found in a chat room about "Janet Jackson" at any given moment are by and large devoted fans. But what if your interest is a fleeting fancy sparked by the majestic presentation of her mammary glands at a certain sporting event? Following the super bowl, hundreds of thousands of people probably performed a search that went something like "janet jackson superbowl" or "janet jackson breasts" or "pop star mammary glands".

Searches are the lightest way people demonstrate their interest in something on the internet, performing several or more searches in a single day. This makes them ideal for detecting when people have shared fleeting interests in a topic. Search Party lets people know how many people are "interested" in a topic at a given time, and then lets them chat together if they are interested.

View the Full Mockup Sequence

Search Party was designed by Seth Nickell (seth@gnome.org). Seth will be happy to work with anyone interested in implementing this feature to make sure we get the user experience right. Please avail of his design skillz!

Technical Notes

One technical key to a good search party implementation is effective "fuzzy matching". We have not specified exactly how fuzzy matching is to occur, but would favor an implementation that included, in addition to basic word-munging (squirrel is roughly the same as squirrels, etc), a simple "semantic" component. The semantic component could be as simple as a probabalistic word-relation table derived from noting proximal occurrence of words in a large source corpus (e.g. "super bowl" is usually found close to "football"). However, don't get too obsessed with this aspect of search party. We'd rather have all the other UI bits working perfectly with a mediocre fuzzy matching implementation than the ultimate academic fuzzy matching library coupled with a crappy search party implementation.

An implementation of Search Party will include an extension (or patch) to either Epiphany or Firefox and a server component. Together they should provide the features/UI described in the complete mockup series and work with Google, Yahoo! and at least one mass market non-web-search-engine that includes a search component such as Ebay, Froogle, Amazon or Netflix.

The server component will aggregate search keywords together, provide information about how many people are searching for similar terms, how many people are in the search party chat for your keyword group, and provide the chat rooms. The sever will need to scale fairly well as search party only becomes tenable if lots of people are using it (and thus the chanes of people looking for info on the same topic at any given time are higher).

If a good argument can be made for it, its ok to use a 3rd party open source component for implementing the chat rooms (e.g. Jabber) with the caveat that this has no impact on the user experience (e.g. no registration is required, search term sharing can be implemented, etc) and it is trivial to setup the server. Overall I suspect that a better user experience can be produced for less work by not using a 3rd party chat server, but I'm not totally adverse to the idea.