Can you read 19th century txt spk?

@guardianstyle on Twitter points to an article by Mark Brown announcing what sounds like a wonderful exhibition the British Library is preparing: Evolving English: One Language, Many Voices (Nov 12, 2010 – Apr 3, 2011).  There’s even a second piece, by Alison Flood.

British Library exhibits are reputed to be large, well-made and almost over-abundant (I’ve only been to one, Taking Liberties, and it was of this style.) If the Guardian is to be believed, usage controversies get a large place in Evolving English, which makes it particularly relevant to this blog. The press has its favourite topics, and one is text-message style abbreviated writing — which is only the latest manifestation of a type of language play that is of course much older. Here is a 19th century example, from the first article:

There will be examples of the linguistic games people played, and a poem from Gleanings From the Harvest-Fields of Literature, published in 1867. In it, 130 years before the arrival of mobile phone texting, Charles C Bombaugh uses phrases such as “I wrote 2 U B 4”. Another verse reads: “He says he loves U 2 X S,/ U R virtuous and Y’s,/ In X L N C U X L/ All others in his i’s.”

I think that modern txt spk would spell XLNs. Also, note the apostrophes in “wise” and “eyes”.

Twitter’s American Airlines i18n mystery

In the Twitter client Tweetdeck, and on my Twitter page itself, I run a search for “i18n”: the frequently used abbreviated form of “internationalization” (or “internationalisation” in BrE). A while ago, I noticed that this search feed contained some odd posts that seemed to have nothing to do with the topic, but originated from accounts belonging to the company American Airlines or from other Twitter users retweeting such posts. Here is a sample screenshot from Tweetdeck:

Odd tweet out: American Airlines in the i18n Twitter search
Odd tweet out: American Airlines in the i18n Twitter search

While all the other tweets have to do with multi-lingual software in some sense, the highlighted one doesn’t. It comes from the user PointsAdvisor and re-posts an American Airlines special offer posted by the corporate account AAirwaves.

I noticed such posts about two weeks ago, both in Tweetdeck and on the Twitter web page, but was stumped until I clicked by chance (or rather, by mistake) on the shortened link inside one of these posts. The AAirwaves account uses the URL (better: URI) shortener Bit.ly, and Tweetdeck will show the original long URI for me to click on (it’s a configurable option, which helps prevent accessing malicious pages). Here is the Tweetdeck link preview for the link in the highlighted tweet above, using data from Bit.ly:

Tweetdeck link preview for AA link
Tweetdeck link preview for AA link

And highlighted, there’s the solution of the riddle: American Airlines uses in the URIs of www.aa.com a path segment (some text enclosed by / characters after the host name) that reads “i18n” — and the Twitter search picks up on this component.

Now on the one hand, this is quite bad URI design on the part of American Airlines, but what’s more interesting is that Twitter’s search engine resolves shortened links and includes the target URIs into the search. I didn’t expect this, as the shortened URIs are posted to Twitter as-is. It could be that the search inclusion is a by-product of resolving and storing the full links for security reasons: to protect against malicious code obscured by a link shortener.

In any event, the effect may be ephemeral. For the last two days, there haven’t been any new AA tweets in my “i18n” search feed on Tweetdeck (which uses the Twitter API). And on the Twitter page, they seem to have disappeared even from the history. I imagine that the Twitter people have to maintain a number of manually created rules to keep search feeds free of accidental spill-over.

This still does not even begin to address the problem of genuinely ambiguous search terms. Wikipedia lists over 20 senses for “FAI” for example, from Fairbanks International Airport to the French term for “ISP” via the Football Association of Ireland, but a Twitter search for the term is overrun by the extremely common Italian verb phrase “fai”. One thing we can expect is for Twitter and similar services to come up with prioritisation and disambiguation options, which, I’d expect, will introduce problems of their own.

Google’s h mystery

A few days ago, my friend Melinda Shore, who knows I’m interested in internationalization, sent me a screenshot from the search bar of her Safari browser. It is a drop-down list of search suggestions provided by Google just after typing the letter h:

Safari search bar Google suggestions
Safari search bar Google suggestions

The top suggestion is a mess:

  1. What does it mean?
  2. Why is it a legitimate search suggestion for the letter h? (If it is.)

Regarding 1., the search suggestion in Firefox is nearly identical, but I cannot reproduce the effect in Google’s own browser Chrome or on the search page directly. In the Safari example, we’re dealing with an odd mix of regular character strings (6.626068, 10, sup, -34), numeric HTML (or XML) entities (×) and raw Unicode-escaped characters that you might find in Python, C or Java source code (\u003C, \u003E). Let’s decode the second and third type of components:

  • \u003C and \u003E simply represent the Unicode code points U+003C and U+003E: the less-than and the greater-than signs < and >.
  • &#215; is U+00D7 MULTIPLICATION SIGN: ×

Putting it together, we get the already much more user-friendly form

6.626068 × 10<sup>-34

or, completing and resolving the HTML: 6.626068 × 10-34.

Once I realized this, my physics training kicked in and the answer to 2. became clearer – Planck’s constant, abbreviated as h,  has the value of 6.62606889 × 10-34 J s (or m2 kg / s). This is not the result of injecting broken text into the search engine results, but a feature of Google’s calculator. Typing “G” into the browser’s search bar also yields similar semi-numeric character salad, while the results for “c“, “e” or “pi” are much more legible.

Still, the entire story raises questions about intent and execution. This is not really an internationalization issue because the form of those physical and mathematical constants is largely invariant by convention. Yet, the tools of internationalization — HTML entities, Unicode code point escapes — have leaked into scientific character display, too. Internationalization is a user interface (usability, user experience) issue [1].

On the execution side, Google got it wrong on several counts, and Apple and Mozilla share some of the blame. Browser search bar drop-down lists don’t allow for superscripts and aren’t sophisticated enough to strip markup, so they display ugly raw HTML. Choosing a numeric entity instead of the character × probably led to its display breaking. And < and > are even in ASCII, so they should display fine, but probably security concerns and their status as reserved HTML characters led to the odd choice of escaping method. All in all, at least one decoding step was not carried out.

More fundamentally, should Google suggest “6.626068 × 10-34 m2 kg / s” when you type a lowercase h? There was a time in my life when I used Planck’s constant daily, and I do use Google’s handy calculator via my browser’s search bar for quick arithmetic and unit conversions. But I think just spitting out the value with no label is going a little too far, and will for more than 99% of users be entirely unexpected: too different from the genuinely useful (for Americans) “hotmail”, “hulu” and “home depot”. Especially considering that for most letters of the alphabet, you could possibly find a scientific constant, function or theorem that starts with it.

Though maybe it is a ploy to spread more science among the people.

[1] It is also a design issue. The two aren’t mutually exclusive.

Google search result for "h"
Google search result for "h" - very different from the broken suggestion

EDIT (2010-08-02): Commenters inside Facebook’s walled garden have remarked that if you actually take up the suggestion and search for it, you get to Planck’s constant. Currently this is partially right, but these things are constantly shifting: In my tests this morning, whether you use the Safari/Firefox  search field or Google’s search page directly, you get a mix of results, the first of which are people wondering about the odd string on SEO forums. A little further down you do get collections of scientific constants, but you have to attentively read the result. Right now, this post is (after less than 12h) number 7 on the results page. None of the pages looks like what you get if you do a Google search for “h” (and hit return) — which is nice and helpful.

Another commenter remarks that for her, the suggestion is now prefaced with “Planck’s constant”, which is a vast improvement.