Its a simple enough idea: voice user interfaces (VUIs) should use ordinarry language, as it is spoken today. They shouldnt be using it in some corrected form to satisfy the nostalgic longings of pedants for some imagined purer form of English.
Its a simple enough idea: voice user interfaces (VUIs) should use ordinarry language, as it is spoken today. They shouldnt be using it in some corrected form to satisfy the nostalgic longings of pedants for some imagined purer form of English. Unfortunately, there is a long tradition of what is known as prescriptive grammar¾ grammar that tries to prescribe the way in which we ought to talk (instead of describe the way we do talk); and it isnt a reliable guide. In fact, prescriptive grammars often try to enforce restrictions that nobody ever respected, not at any time back to the Middle Ages. There are grammars out there that contain rules that were never genuine rules of English grammar at all. That fact holds some real dangers for dialog designers. Suppose you encountered a book on Unix that said some features of current OS-X or Solaris are incorrect and recommended early VAX Unix commands, or commands that the author thought would be better than the real ones. Suppose you read ls l /usr, with the option before the directory name, is a barbarism; ls /usr -l is the preferable form. Youd be annoyed that the recommended command didnt work; youd be positively outraged when you found out that ls /usr l had never worked on any variety of Unix. But users of English grammar and usage books put up with this sort of abuse every day. Ten famous cases are listed in the sidebar (see Fear no more). Consider singular they first; it probably gets more attention than any of the others these days. It is really useful: English needs a device that gets around the lack of singular third person pronouns that are truly sex-neutral. There is a myth more like a flat lie, actually that the masculine pronoun he may be used sex-neutrally. Hogwash. On the back of the paperback 4th edition of the famous little book The Elements of Style by William Strunk and E. B. White (Boston: Allyn and Bacon, 2000) is a promotional quote in small capitals from the journal Telephone Engineer and Management. It says:
ITS HARD TO IMAGINE AN ENGINEER OR A MANAGER WHO DOESNT NEED TO EXPRESS HIMSELF IN ENGLISH PROSE AS PART OF HIS JOB. ITS ALSO HARD TO IMAGINE A WRITER WHO WILL NOT BE IMPROVED BY A LIBERAL APPLICATION OF The Elements of Style.
Its also hard to imagine anyone who doesnt see that the first sentence is making the assumption that all telephone engineers and telephone-industry managers are male. If phrases like express himself or part of his job could really be taken as allowing for female managers expressing themselves as part of their job, it should be normal to say things like this:
Whether we hire a man or a woman, when he arrives he will be treated like everyone else.
Anyone who fails to see that this is a joke simply doesnt know how English works. The normal way of handling such situations in English is to use they, especially with indefinite antecedents like everyone, nobody, who, an engineer, a man or a woman, etc.:
Whether we hire a man or a woman, when they arrive they will be treated like everyone else.
Is there any reason relating to grammar or proper usage that makes such sentences bad? The answer is no. We find singular they in the 16th century with the writings of Shakespeare; and we find it in all subsequent centuries. It was particularly common in Jane Austen: Who makes you their confidant? says a character in Emma. Its occurrence is commonplace in good writing. Microsoft Word will tell you otherwise: it puts the green wavy line under singular they or them or their wherever it spots them. It will tell you to change them to he or she, him or her, his or her. Just click Ignore. Word has very conservative ideas about grammar. What were saying is that a dialog thats been run through Words grammar checker could come out worse, not better. Singular they is becoming even more frequent in modern conversational speech. Someone on their way to work can have an email read to them, says David Weiden of Tellme Networks in a recent interview with this magazine; that is how we talk. People are even beginning to say things like Ill ask my partner about it and see what they think (completely concealing the matter of whether company policy on same-sex domestic partners is relevant to them). It would not be a worthy gesture toward maintenance of linguistic standards to keep singular they out of VUI dialogs. It would be a mistake, and would make them sound less natural. The same is true for many other aspects of grammar. Take, for example, the famous piece of nonsense saying that a preposition is a word you shouldnt end a sentence with. The first sentences in English that have prepositions at the ends of clauses occur over seven hundred years ago. Since then, there has never been a time when such sentences did not occur. Then in 1672 John Dryden wrote an essay on style in which (for unknown reasons) he cast aspersions on final prepositions, and an urban legend of grammar was born. Ever since then there have been some people who imagined that Drydens delusional prejudice had a real basis to it. It has none. English is not a language like Spanish, in which a preposition is always immediately followed by its object noun phrase. It is (and has been for the best part of a thousand years) a language like Icelandic, which allows a preposition and its object to be separated. Question words yield the clearest examples. We say: What did you do that for? and Who are you looking at?; the frequency of real-life humans saying For what did you do that? or At whom are you looking? is roughly zero. So we applaud Tellme, whose main menu greeting includes the sentence:
Here are all the categories you can choose from.
If they had corrected this to Here are all the categories from which you can choose would have made it less natural, less like a normal person speaking. Similarly, BeVocals service tells callers to:
Just name the sport you want to hear about.
But somehow the close juxtaposition of prepositions in BeVocals main menu To log in, say Go to log in. takes the point a little too far, clipping it into robotspeak. A lot of people find it a bit harder to talk naturally on the telephone. Their residual insecurity about the way they speak English is amplified by the anxiety, and they reach for forms of words they have never been comfortable with and do not normally use. So some stranger calls up and asks for Angela, and Angela says This is she. But when Angela is asked at a party to point out Penelope, she doesnt point to Penelope and say, That is she; she says Thats her, over there. Thats the way we normally talk. Linguists who have done the necessary transcription and counting say that only in one or two cases out of a hundred in real conversation will the pronoun after is be she rather than her. Many of those are probably uses of This is she or This is he, which are almost fixed phrases for telephone use, limited to the context of confirming ones identity on the phone. Much the same holds for I and me, of course. A cartoon we once saw had it right: a woman inside her house was saying suspiciously through the closed front door, I dont know anyone who says It is I. When someone we know calls out Who is it?, our natural response is to say Its me. Some instances of anxiety-driven language of this sort are beginning to be institutionalized. How may I direct your call? is one. Another, more interesting, is:
Whom should I say is calling?
First, whom is so rare in spoken interrogatives that the Longman Corpus of Spoken and Written English does not contain a single example. It is not just a little stiff and starchy; its as dead as a dead parrot. And second, the really strange thing about Whom should I say is calling? is that its use of whom is arguably an error anyway. If we clear should I say out of the way, we get Who is calling?, not Whom is calling?. The word who is logically the subject of is calling, not the object, so whom is not correct here in standard grammatical terms. What makes whom such a problematic feature of modern English is that it is so close to disappearance. We scarcely know how to use it at all. And that means we shouldnt be trying to make the voices in our VUIs use it. It is the job of VUI designers to minimize users anxiety and unnaturalness. And it is the role of clients (such as airlines and financial institutions) to let the designers do their job freely and well, without meddlesome direction or demands based in half-recalled rules about correct English and formal versus casual usage. They should start by getting a serious look at what is grammatical in English and what is not. And that is not so easy: the standardly accepted views about such things are very often wrong. REFERENCES YOU CAN TRUST
There are a few reference works that may be of help to a dialog designer. For an up-to-date, comprehensive survey of English, the latest and most complete is The Cambridge Grammar of the English Language
by Rodney Huddleston and Geoffrey K. Pullum (New York: Cambridge University Press, 2002), an 1862-page compilation combining insights from modern linguistics and traditional scholarship without presupposing a linguistics background. Among usage manuals advising on points of contention in English, the best is Merriam-Websters Dictionary
of English Usage (Springfield, MA: Merriam-Webster, 1994). And of course, for anyone doing careful work on language, top-quality dictionaries should be available. For American English, the two best are Websters Third New International Dictionary
(Merriam-Webster), which was controversial for its lack of prescriptivism when it came out in 1961, but is now completely standard, and the 4th edition of The American Heritage Dictionary
(Houghton Miflin), which has extremely useful usage notes by Stanford linguist Geoff Nunberg. The court of last resort for grammatical issues covering the whole history of English is, as always, the 2nd edition of the Oxford English Dictionary
, except that on pronunciation issues it should be checked against the new edition of the Longman Pronunciation Dictionary
by J. C. Wells (Longman, 2000), which covers both British and American English.
Dr. Caroline Henton is CTO of Talknowledgy.com. Dr. Henton can be reached at firstname.lastname@example.org
or (831) 457-0402. Dr. Geoffrey K. Pullum is professor of linguistics at the University of California, Santa Cruz, and a coauthor of The Cambridge Grammar of the English Language. He is email@example.com
by email and can be reached on (831) 459-2555.