Hacker News Reader

Comment by tptacek | Hacker News Reader

Google's job is to show the search results that most closely match what users are actually looking for. In 2015, far more Google users are looking for Rand Paul than Paul Rand. Would that it weren't so, but it is. Paul Rand's logos might touch millions of people every day, but Rand himself is not relevant to the majority of Google users.

View on HN

I'd say that Google's job is to show the search results that _each_ user is actually looking for. (Not an aggregate user with 2.5 children.)

Come to think of it, doesn't Google aspire to personalized search? But in any case, at least I'd like old-fashioned quoted search -- i.e. I want "this phrase" exactly -- to give useful results.

I didn't think phrase matching (quotes) was ever broken? I know Google dropped the "+" filter a while back, but I still continue to use phrase matching quite heavily, and it seems to still work as it always has.

The string matching is fuzzy though. It is case insensitive and it ignores at least some non-alphanumeric characters.

For example, "twenty-four" will return results about the TV show "Twenty Four" as well as results for "twenty four" and "twenty-four".

There's some other maddening exceptions for those who would prefer exact string matching, but I can't remember them all and it's changed over the years too.

I think it does. After clicking this article, my top four search results for Paul Rand (without quotes) were the designer.

I enjoyed this piece as sort of a tongue-in-cheek way to justify this article's existence -- if more people read or link to this article, it might end up showing higher in Google's results :)

Moreover, the top results I see -- under a paid ad for Rand Paul -- are paul-rand.com, his wikipedia page, and a right-hand panel about Paul Rand. Is it so terrible that some other of the ten blue links are for people who confuse the two? Is there another vital Paul Rand result Google is missing?

I get:

  (Sidebar: Paul Rand)
  Paul Rand
  Paul Rand
  Rand Paul
  Rand Paul
  (News: Rand Paul)
  Paul Rand
  Paul Rand
  Paul Rand
  Paul Rand
  Rand Paul
  Paul Rand
  Rand Paul

So... 8/12 or 75% of the first page of results are Paul Rand. Plus a thing underneath the sidebar saying "See results for Rand Paul".

That's seriously not a disservice.

Author. One point I want to add here: If you're looking for news articles about Paul Rand, you'll be sorely disappointed.

Even in quotes—where the exact phrase is supposed to show up—most of the articles are about Rand Paul. That points to algorithmic issues.

But at the same time, it's worth noting that the other examples of this issue I cite—in the case of Siri and Bing—are even more troubling, because they seem to drown out the figure entirely. Based on a search of Bing in incognito mode, no searches, I found that "paul rand" in quotes brought up 80 percent Rand Paul. Siri only brings up one result for "Who is Paul Rand," and it brings up the Wikipedia page for Rand Paul.

I realize people are going to disagree about the thesis of the article, but I think my goal here was to point out that problems like these exist—and are likely to become more common over time. Thanks for engaging.

The problem is that even when you put "Paul Rand" in quotes, the search results still show "Rand Paul" half the time.

This is ever the problem of general purpose data handling systems. They optimize for popularity not for some kind of domain desirability.

That's what I came to say. The job of Google is to interpret and return relevant results. If you're looking for the designer, then you'll need to tack on 'designer' to the search.

I was watching one of those 'investor pitch TV shows' on TV, and a young girl was pitching her site, which was named using a misspelled word (presumably to secure a domain name otherwise unavailable). If I was on that panel, I would have told her to go back to the start and find a new name and branding. She even admitted on the show that she wasn't turning up in results because of auto-correct spelling.

These days, you can't pick company names or brands that closely match something else. You are better off starting with a completely new word than trying to misspell something slightly.

Reply on news.ycombinator.com