This utterance tag records certain features of an utterance unit such as whether it was interpretable. Unlike the other categories where a choice needs to be made between options, here the features are independent of each other. Since these features only mark exceptional circumstances, most utterance units have no features marked in this classification. The features marked are
Uninterpretable
The utterance unit is not comprehensible. It may or may not be Self-talk depending on how it sounds. These utterances are rare and are usually word fragments although sometimes bad grammar and bad pronunciation/typing can make an utterance impossible to understand. An utterance may also be semantically ill-formed although syntactically perfect.
Abandoned
Sometimes a speaker will make an error in their utterance or change their mind about what they are going to say, and simply abandon their utterance. Such fragments are marked as abandoned only if they provide no content to the dialog, i.e., the import of the dialog would not change if these utterance units were removed. If the utterance does provide content to the dialog, it should not be marked abandoned and should be annotated as usual. In the example below, the abandoned utterance ``so I pick up'' does not provide content to the dialog and so is labeled abandoned.
Abandoned utt1: u: so I pick up ..
utt2: can I take oranges um on tankers from Corning
Note that just because an utterance is broken off does not mean it is abandoned as defined here. If it contributes content to the dialogue, then it is not marked as abandoned. For instance, consider
utt1: s: You'll need two boxcars. utt2: u: okay, and we-
Utt2 is not marked as abandoned as it serves at least to acknowledge utt1, and possibly to accept it. Thus it contributes content to the dialogue.
Self-talk
The utterance unit consists of one speaker talking to him or herself. We assume this does not occur in typed dialogs. This is impossible to detect without listening to the actual speech. Just because an utterance is labeled as Self-talk doesn't mean that the other agent can't hear it, or can't use the information. It simply means that the speaker appears to not be intending to communicate what is being said.
Discussion
While most of the categories are fairly straightforward, more examples and explanation are needed concerning what makes an utterance Abandoned. If a complete clause is uttered then it should not be considered abandoned even if some extension to the clause is broken off. In the dialog excerpt below, ``it's four hours'' is a clause. u never gets to start the new clause introduced by ``and'' but ``it's four hours and'' should not be labeled abandoned, since the content of ``it's four hours'' has been communicated in the dialog. ``how long'' should be labeled Abandoned because it does not play a part in the dialog.
utt1: u: mm <click> okay Reassert utt2: four hours from Avon to Bath Action-directive utt3: and then I guess attach that to the boxcar to Corning Reassert utt4: it's four hours and Abandoned utt5: how long Info-request utt6: it is two hours from Bath to Corning
In this example, ``um then go to Corn-'' would be labeled abandoned even though it was broken by s instead of u.
Action-directive utt1: u: then take one more boxcar from Bath to Corning
<click> fill up with oranges
Abandoned utt2: um <sil> then go to [Corn-](1)
Reject(utt1) utt3: s: [there are no](1) boxcars available with Bath
unless you're talking about the ones that are
filled with bananas