Friday 13 July 2012

Three reasons why people search sucks

The Tattoo Bible not by Alex Guest

I am not a holidaying establishment.

The Alex Guesthouse. Google and other search engines are inept when it comes to distinguishing between a holidaying establishment and a person.

I was not portrayed by Andie MacDowell. 

Alex (guest star Andie MacDowell)... They are rubbish at detecting punctuation that fundamentally alters the meaning of the result.

I did not write The Tattoo Bible.

And useless at evaluating two or more people sharing the same name: it is a different Alex Guest who can take credit for The Tattoo Bible.

There have been attempts to deal with this but they are all equally useless.

You think it's just me and my slightly quirky name?

A friend of mine, a doctor, shares her name with a writer, a model, a wedding photographer and a few other doctors. I can tell the difference between each of these. I could manually group the results by individual and serve up the results as clusters relating to each one. As a human, it is easy for me to tell the difference between an oncologist based in Ireland, and a writer of trashy novels, whose agent is in New York City.

Six steps to fix people search

  • First, consider the links coming in and out of the page and especially links between those pages.
  • Next, look for clues such as title: Dr, Mrs...
  • Go further and consider some of the words that appear specifically in some results and not others. My doctor friend practices in a different field to the other doctors with the same name. Indeed, it appears that each one of them - on a quick analysis - have unrelated specialties.
  • What geographical locations are referenced?
  • Some names, like Alex, are given to both men and women. Look for gender clues in pronouns.
  • Consider that some names are words with everyday usage: Guest, Brown, Parkinson. Alex (guest star...) should be trivial to detect as it has punctuation between Alex and guest; guest is all lower case; and guest star is a common word pairing.

Finally, the word Tattoo, an indelible mark on the skin made by inserting pigments in punctures, is clearly not the same thing as Zattoo, the live TV-streaming business for which I was the UK Country Manager.

It should not be difficult for a machine to cluster search results for individuals, yet, as far as I know, there is currently no solution that does this adequately. If you can write code and are irked by this problem, get in touch. Please.