next up left folder
Next: Backward Looking Function Up: Forward Looking Function Previous: Committing-speaker-future-action (Influence-on-speaker)


Note that the Forward Looking Function is a characterization of what effect the utterance has on the dialogue, even though the actual form of the sentence might look like something else. Each utterance could have multiple tags in this aspect depending on how many functions it simultaneously performs. As an example, in the right context, an utterance such as ``There is an engine at Avon'' could be both an Assert and an Open-option (stating the possibility of using that engine to move some cargo).

Other examples arise in dialogs where the agents are performing a task together. For instance, if the two agents are trying to agree on how to furnish a room, a suggestion like ``Why don't we use the red sofa for the living room'' is both a Offer (roughly, if you agree then I will use the sofa) and an Open-option (roughly, I state the possibility of you using that sofa). If this utterance is accepted by the hearer, then the net effect is that both participants are committed to using the red sofa.

Note that the previous version of this scheme had a separate category for conventional acts like thanking and apologizing. These acts now fall into different categories based on the tests presented. For instance, an utterance like ``I'm sorry'' is an Assert as it makes a claim about the world, which you can see by observing that it could be followed by the response ``No you're not''. The utterance ``Thank you'', on the other hand, cannot be responded to in such a way and only falls into the explicit performative category (observe that ``I hereby thank you'' is fine).

The Info-request and Influencing-addressee-future-action (Influence-on-listener) dimensions are similar in that they both apply to suggests and requests (Info-requests request communicative actions and Influencing-addressee-future-action utterances request non-communicative action). Table 1 gives examples of how various questions, imperatives, and indirect requests should be labeled in these aspects.

Table 1: Examples of how to label Info-request and Influence-on-listener

Short utterances such as ``okay'' and ``yes'' can have many different interpretations depending on how they are being used. In fact, they can be Asserts, Commits and even Other-forward-functions. If the utterance conveys information in answering a question, then the utterance is only an Assert, as in the dialogue

Request-info  Utt: s: is there a train at Chicago?
  Assert        Utt: u: yes.

On the other hand, if the utterance commits the speaker to a certain action, say in response to a request or invitation, then it should only be labeled as a Commit

Action-directive Utt: u: tell me if the route gets too long
  Commit           Utt: s: okay

Cases where the speaker utters an ``okay'' to accept a request and then performs the act requested are also marked as a commit, even though the action to which the speaker is committing is performed immediately after, e.g.,

Action-directive  Utt: u: can you tell me the time?
  Commit            Utt: s: okay.
  Assert            Utt:    three o'clock.

Utterances such as ``okay'' can also be used not only to convey information but to manage the dialog, possibly to hold the turn, or to signal the introduction of a new topic. In these cases, the Forward Function of these utterances is not directly captured in the annotation scheme, and they should be marked as Other-forward-function. This category serves as a place holder for groups that want to analyze such utterances. Some communicative functions for these utterances are also captured in the annotation of Backward Looking Function that is described next.

next up left folder
Next: Backward Looking Function Up: Forward Looking Function Previous: Committing-speaker-future-action (Influence-on-speaker)

Mark Core
Mon Sep 22 20:05:25 EDT 1997