10.3 Information Structure


For the purposes of this book, let information structure simply refer to the placement of contextually determined "old" versus "new" information, whether in recorded messages or in natural conversation. For example, consider the following interaction between one of the authors (JG) and a local directory assistance service.

(25)

LIVE 411 OPERATOR:

Hi, this is Joan. What city?

CALLER:

Mountain View.

LIVE 411 OPERATOR:

All right?

CALLER:

CostPlus.

RECORDING:

The number you requested, five, five, five, nine, six, one, six, zero, six, six, can be automatically dialed by pressing one now. An additional charge will apply.


Except for the prosody of the phone number (see Chapter 11), there is nothing strange about the recorded message, as far as individual words and sentences go. Yet the way the system gives the requested number is nonetheless uncomfortable. Although we have not conducted a formal study, anecdotally people say, "I have to hang up in the middle of the sentence if I want to remember the number." The problem has to do with the way this recording figures in context specifically, the way that the "old" versus "new"information has been laid out for the listener. From the point of view of most callers, the new piece of information, or focus, should be the requested phone number (555 961 6066), but this information is buried in the part of the sentence that is conventionally reserved for old, or nonfocal, information. Such placement actually encourages us to forget what the number is, even though that was the purpose of the call to begin with.

VUI designers and prompt writers should be aware that there is a neutral (default) position of focus in English, as determined by what is called the End-Focus Principle (Quirk and Greenbaum 1973). That is, native speakers of English naturally place new or focal information at or near the end of the sentence, and this is where native hearers of English naturally expect to retrieve it. An example of a conversational message that not only complies with the End-Focus Principle but also facilitates short-term memory would be, "Here's the number: 555 961 6066. For automatic dialing, press one . . ."

The way people naturally structure information is highly context-sensitive. In the case of the directory assistance recording, it is context that makes the positioning of the phone number inappropriate. It is easy to imagine some other context where a phone number would occupy this position appropriately, as in frequently heard messages of the type, "The number you have dialed, 444-4444, is not in service. Please check the listing and dial again." In this context, the already dialed phone number is old information, and so its nonfocal placement in the middle of the sentence is justified.

To be technically accurate, end focus is given to the last open-class item or proper noun in a clause (Quirk and Greenbaum 1973).[3] Open-class items are nouns, adjectives, verbs, and adverbs. These categories are called "open" because they are open to new members (witness the creation of such verbs as "downsize," "e-mail," and "dis"). In contrast, closed-class items are articles, demonstratives, prepositions, and conjunctions (these are categories that reject new members; no one can invent alternatives to words such as "the," "that," "of," or "and"). In the sentence "Sean Connery was born in Scotland," the last open-class item is the noun "Scotland." By default, it is the focus, the new piece of information in this sentence. In contrast, "Sean Connery" is the topic (subject) of the sentence, or the old piece of information on which the speaker makes some comment. Old information is generally placed in the subject, whereas new information is generally housed in the predicate.

[3] Focus can occur at earlier points in the sentence, but this requires the use of special, "contrastive" stress and intonation, as in "Sean Connery was born in Scotland," which would answer the question of who was born in Scotland.

Because prompts are sometimes written out of context, it is easy to find examples that violate principles of information structure, which is context-sensitive. For example, in a certain application that recognizes dates, there are special messages associated with months that have only thirty days. These messages are played only when the recognizer returns, for example, "June thirty-first," or "February twenty-ninth" in a nonleap year, with high confidence. To comply with the End-Focus Principle, we would expect such messages to be worded as in (26), but in fact, they are worded as in (27).

graphics/sound_icon.gif

(26)

CALLER:

Make it for June thirty-first.

SYSTEM:

Actually, June has only thirty days.


graphics/sound_icon.gif

(27)

CALLER:

Make it for June thirty-first.

SYSTEM:

There are only thirty days in June.


The wording in (26) appropriately casts "June" as the subject of the sentence. In this position, it assumes the role of topic, on which some new comment will be made. The new information is the "only thirty days" part, which is appropriately cast in the predicate, where it receives end focus.

It seems plausible that the End-Focus Principle has something to do with the short-term memory phenomenon known as the recency effect, as discussed in Chapter 9. Elements that come at the end are more prominent and easier to remember precisely because they are the last thing one hears. To exploit these linguistic and psychological principles, it makes sense to structure directions so that the novel element, such as which key callers should press or which voice command they should say, follows the topic, which is the objective or task. Preferably, the touchtone or voice command should be placed at the end of the prompt.

Careless handling of focal information can lead to ambiguity and wrong inferences. For example, a certain application allows callers to choose from a list of song titles and download the musical selection to a cell phone. To select a song, the caller is supposed to say, "That one!" upon hearing the desired title. In the sample interaction in (28), the system prompts the caller to say whether or not it got the selection right, simultaneously informing the caller about the cost:

(28)

[System plays list of Britney Spears selections] "Oops, I Did It Again" <pause>

CALLER:

That one!

SYSTEM:

Would you like to buy the song "Oops, I Did It Again" for $4.99?


What is being sought here? Is it confirmation of the song selection ("Oops, I Did It Again") or of the price ($4.99)? Context and world knowledge likely indicate the former, but the information structure, with the price in the end-focus position, suggests the latter, as if the price were open for negotiation. In other words, the prompt is ambiguous, permitting two valid interpretations. As a legitimate response, a haggle-prone caller might legitimately reply, "No, but I'll give you a buck fifty for it." The prompt can be recast unambiguously as something like this: "Good choice <pause>'Oops, I Did It Again' is only $4.99. Do you want to go ahead and buy it?" Here, the price has been removed from the scope of interrogation, and the prompt has only one possible interpretation.

In light of the End-Focus Principle, we should have a renewed appreciation for the communicative function of the passive voice in conversation as well as in prompts. Consider the active and passive sentences in Figures 10-1 and 10-2, respectively. In both cases, "Leif Erikson" is the agent (that is, the do-er), and "North America" is the patient (the affected element) of the verb "discover."

Figure 10-1. In the active voice, the agent is the old or topical, and the patient is the new or focal.

graphics/10fig01.gif

Figure 10-2. In the passive voice, the patient is the old or topical, and the agent is the new or focal.

graphics/10fig02.gif

These examples illustrate a fundamental difference in the use of the active and the passive voice. In the active voice, the agent is the topic, or old information, and the patient is the new information, or focus. In the passive voice, these roles are reversed: The patient is the topic (or old information), and the agent is the focus (or new information). So the passive voice is useful because it allows the speaker to reverse the usual order of agent and patient, thereby reversing their default informational status.

The choice of the passive voice is more than appropriate it seems to be required in the recorded announcement in (29).

(29)

This program has been made possible by a grant from the Ford Foundation.


If we attempt to rewrite the sentence to "avoid the passive voice," as suggested in some texts on writing style, the result is strangely off-balance: "A grant from the Ford Foundation has made this program possible." In the context in which we would hear (29), the piece of information that requires focus is "the Ford Foundation." What is old, topical information is "this program," the one we've been watching for the past hour. The activized version of (29) strikes the ear as off-balance precisely because it flouts the End-Focus Principle.

The passive voice is used appropriately in the following two examples from a telecommunications VUI. The company that owns the application wanted to make it very clear to callers who exactly would be handling their request. Some requests would be handled by the new, automated speech system, whereas others would be routed to customer service representatives, as before. The exchanges in (30) and (31) illustrate these two dialog possibilities.

(30)

CALLER:

Copy of my bill.

SYSTEM:

Your request for a copy of your bill will be handled by our Automated Express Service.


(31)

CALLER:

Clarification, please.

SYSTEM:

Your request for clarification of an item on your bill will be handled by the next available representative.


In the system's responses in (30) and (31), the old information "your request for a copy of your bill" and "your request for clarification," respectively is maintained as old information in the subject slot, and the new pieces of information "our automated express service" and "next available representative" are accordingly given end focus. If the prompt writer had put these sentences in the active voice to satisfy his or her grammar-checking software, the result would have been stilted, context-insensitive messages such as, "Our automated express service will handle your request for a copy of your bill."

This discussion of information structure reveals the significance and the implications of speech's linear organization. The prompt writer must consider the role of context in judiciously sequencing elements for a familiar, comprehensible listening experience.



Voice User Interface Design 2004
Voice User Interface Design 2004
ISBN: 321185765
EAN: N/A
Year: 2005
Pages: 117

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net