Mon 7 Dec 2009
Over the weekend, I planned to go north to Marin County for an art opening. Since that event didn’t start until 5 p.m., I thought I might get a little holiday shopping in, and I remembered a company in Marin that makes products (such as children’s bibs, aprons, lunch sacks, and tablecloths) out of cloth in bright patterns that’s coated and easy to clean. I hadn’t visited the physical location before, but didn’t I read something in the newspaper about a warehouse sale happening around this weekend?
Now, I couldn’t dredge up the exact name of the company, but I knew the pattern: [female proper name] the [animal name]. I actually knew more: The girl’s name was 2 syllables, like “Emma” or “Anna.” And the animal was something aquatic or amphibious: Frog but not horned toad, like salamander but probably only 2 syllables also. So I knew many parts of the company’s name but not the whole.
Girl’s name + aquatic animal
(2 syllables) the (2 syllables)
Search can’t take input like this and turn it into results — yet. Humans, however, are known to categorize and associate items by sound (including rhythmic structures) as well as by meaning. (We also make visual confusions based on spelling or the shapes of letters, but that’s not relevant here.) Current search algorithms are based on keywords, patterns of misspelling, and some more complicated semantic information, including polysemy (the word bank, for example, refers to “financial institution” but includes meanings for both the organization and its physical site). Success of one search engine over another may rely on avoiding homonyms while tracking polysemes: A searcher gives cues to help the search engine distinguish the word bank as the shore of a river or stream from the same word in its other senses.
Now… what was the company I was looking for? And how did I find it? I got off the freeway, wandered into Whole Foods in San Rafael, and immediately identified an apron from this company.
Was it “Anna the Salmon,” ”Emma the Clamshell,” or ”Mimi the Sardine”?
Notice we’ve also got vowel harmony here too! Lots of cues, but not the sort I could have used today with Google or Bing or [insert favorite search engine]. Even among humans who hadn’t heard the name before, I could have used the familiar charades gesture for “2 syllables.” Seems like I was pretty close in my recall of this name. But “pretty close” doesn’t work for web searches today. Close is only good in horseshoes.
The company’s factory outlet was not open that day; the warehouse sale was happening elsewhere. Now I’ve learned they’re on Twitter, and I’ve marked the name with this long rant, so I won’t forget my friends at www.mimithesardine.com.