Unapplied Linguistics


I bought a train ticket online from Virgin Trains recently, to get me to St. Neots, nearest station to the site of this conference, where I’m speaking to an association of freelance editors. The follow-up email from Virgin Trains surprised me. The subject line said: “Your St. Neots journey, your way.”

Your St. Neots journey is a well-formed English noun phrase using the proper name St. Neots as an attributive modifier of the noun journey. Trivial to program: They simply had to take Your _____ journey and substitute into the blank the name of the destination I had picked. Replacement of a variable call by the current value of the variable. Elementary. But this obvious little trick represents a relative high point of natural-language computer communication with human beings. Nearly every machine that speaks to me or messages me in a working day is unutterably stupid, down at the D-minus level, whereas the idiomatically worded robot email from Virgin Trains merits a C, even C+.

Elsewhere, the usual chaos reigns: sloppy programming, basic grammar errors, idiotic assumptions about the user’s mental state and needs. Everywhere that machines communicate with me, I see signs of utter incompetence.

• A database search for a company yields just one result, and the website tells me: “There are 1 companies at this address.” Even I could fix the programming to make it write that grammatically.

• German in-car GPS systems, according to my friend Lieselotte, make voice announcements mentioning streets and roads, but do not have even the rudimentary gender classification of nouns that would be necessary to distinguish Weg (“way”), which is masculine, from Strasse (“street”), which is feminine, and so on. Not even the most basic details of the gender classification for nouns, determinatives, and some adjectives have been programmed in.

• On the new Xerox copier-printer near my office you have to log in to release a print job, and once you have released it and it’s being printed, the communication screen announces: “No document found.” Of course a document was found: it was just passed to the printing routine. How sensible is that?

• I call up my bank and a machine answers, telling me it will speed things up if I key in the 14 digits of my account number, which I laboriously do. Two minutes later I’m finally talking to a human. He asks me for my account number again. As usual, I demand to know why I was put to the trouble of punching it in on the keypad if it was going to have no effect, and he tells me that although it was entered into the system he cannot see it on the screen he has to consult. So glad I bothered.

• I arrive at the airport, check in, go through electronic passport control and security, and walk to the gate. The weather app, the time zone app, and a message from the local cellphone service all show that they know exactly where my phone is. And while I am standing near the gate, an audible signal alerts me to a new text. It’s the airline, telling me that I have a flight soon, so I should go to the airport. Excuse me?

• I receive a new credit card in the mail. I log in and activate it online. Two days later, I get a text on my phone from the bank:

We’ve sent you a new Credit Card. Please sign the back & choose an activation option:
1) Log-on to Online banking, click Activate new card
2) Call 08003281370

Someone didn’t bother to put into the program a routine that checks whether I have already logged in and clicked “Activate new card.”

And so on, every single day. You must have experiences like this. Maybe you dismiss them as minor amusements, if you notice them at all. But they are indicators of something important, something I have commented on here before, though nobody pays attention so I’m saying it again.

Although some of the problems will relate to failures of database integration, many of them suggest that the programmers who are assigned to write programs controlling machine-to-human interaction are writing crap software. They don’t care, and their managers don’t care that they don’t care. We get software written by Wally the slacker rather than Dilbert the conscientious engineer, software that serves up messages that are morphologically, syntactically, or semantically aberrant — and often completely inappropriate to the context. Part of the problem is that linguistics is not yet being applied in the key arena where it could most profitably be applied: the artificial-intelligence business.

They keep telling us that AI will revolutionize our future. Revolution? The software governing our everyday interaction with machines is not even minimally fit for its trivial purposes.

Return to Top