tag:blogger.com,1999:blog-95544842011-04-21T14:37:56.238-07:00Kendall Willetshttp://www.blogger.com/profile/02028216934590585245noreply@blogger.comBlogger21100tag:blogger.com,1999:blog-9554484.post-1103002068049320322004-12-13T21:15:00.000-08:002004-12-13T21:27:48.050-08:00That guy had a certain look I couldn't place for a while...moustache, shaved head, cop. Hmmm...then it hit me: the gay police captain in Mad Max! I guess I know what Bush liked about him now. <br /> <br />Followup to Google completion: finally found <a href=http://blog4.bitflux.ch/wiki/LiveSearch>Livesearch</a> at bitflux.de. This looks like it will do what I want it to do; I just have to figure out what the backend should supply and write an fcgi to do it. Maybe it'll be a type-and-go human genome searcher, or some other corpus. <br /> <br /><div class="blogger-post-footer"><img width='1' height='1' src='https://blogger.googleusercontent.com/tracker/9554484-110300206804932032?l=kendallwillets.blogspot.com' alt='' /></div>Kendall Willetshttp://www.blogger.com/profile/02028216934590585245noreply@blogger.com4tag:blogger.com,1999:blog-9554484.post-1102706750122024102004-12-10T10:55:00.000-08:002004-12-10T11:25:50.123-08:00After reading of <a href=http://insidegoogle.blogspot.com/2004/12/google-suggest.html>Google Query Completion</a> I realized that they had come up with the first query-as-you-type for web search. I had tried to write something like this a few months ago and given up after the first cryptic Javascript errors. <br /> <br />It's a pity they use this just for query completion, because they could use it for incremental search as well, with a few restrictions. <br /> <br />As I suspected, the <a href=http://www.google.com/webhp?complete=1&hl=en>suggest page</a> is driven by <a href=http://www.google.com/ac.js> Javascript spaghetti code</a> which bangs away on an http interface after each keystroke. The spaghetti fetches a url like this: <a href=http://www.google.com/complete/search?hl=en&js=true&qu=chicken>http://www.google.com/complete/search?hl=en&js=true&qu=chicken</a> (here "chicken" is the string just typed). The return appears to be a Javascript subroutine wrapped around the top 10 results. <br /> <br />The back end appears to be just a trie of queries, with counts on each node. Only prefix matches appear to work, so, for example, one can find "endives au jambon", but nothing matching "au jambon". The data seems to be static; I don't see the counts changing when I reload the url. <br /> <br /><div class="blogger-post-footer"><img width='1' height='1' src='https://blogger.googleusercontent.com/tracker/9554484-110270675012202410?l=kendallwillets.blogspot.com' alt='' /></div>Kendall Willetshttp://www.blogger.com/profile/02028216934590585245noreply@blogger.com2