next up left folder
Next: Appendix: Outstanding Issues Up: Draft of DAMSL: Dialog Previous: Tagging Cue Words

Mechanics of Annotating

 

A new dialog annotation tool (dat) is available on our web site:

http://www.cs.rochester.edu/research/trains/annotation
  

The site contains instructions on how to setup and run dat. When you start dat, you can give it a dialog file to open. You can also chose Open from the File menu to open a dialog. dat expects files to be in a format called DAMSL (Dialog Act Markup in Several Layers) which is a variant of SGML tailored to this application. The dat tool can convert raw text files into DAMSL and the man pages of dat contain more information about this format.

The sets of tags discussed previously can be applied through menus, buttons, and text fields (in the bottom half of the dat window) once an utterance or segment is selected. dat allows you to select utterances by clicking on them. Clicking on another utterance while holding the shift key allows multiple utterances to be selected. Once you have selected one or more utterances, you may select various Utterance-Tags for the utterance(s). In addition, if the utterance(s) are a response then choose the appropriate Backward Function labels and indicate what was responded to, using the Response-to field. In this field, type the names of the utterances that a response pertains to, and separate names by dashes (NOT spaces). In the example below, you should type ``utt1-utt2'' in the Response-to field of utt3 since it accepts both utt1 and utt2.

T1  utt1 u: take the engine to Corning
  |   utt2 u: there will be a tanker there and we will take that to Elmira
  T2  utt3 s: okay
  

The Communicative-Status tags are implemented as menus near the middle of the screen. If a tag applies then select ``yes'' or ``maybe'' from its menu depending on how certain you are. All of the other dialog tags have question-mark buttons next to them allowing uncertainty to be encoded.

Once you are done annotating an utterance or a set of utterances, click the ``Apply'' button. Click ``Reset'' to eliminate any changes you made since you last clicked ``Apply''.

We encourage you to listen to any speech associated with each utterance. To do so, select the utterance unit and click the ``Play Speech'' button.

Note, tagging a series of utterances with a label such as Assert means that you have a series of Asserts. If you would rather a series of utterances be considered one unit then form a segment from the utterances and label the segment. To form a segment, first select a series of adjacent utterances and then choose Define Segment from the Group menu. You can click on this new segment and label it just like an utterance.

When you wish to stop annotating, select Save As from the File menu. The dialog box is similar to the one you saw when opening a file. Edit the filename to create a new file to save your changes in. Selecting Save afterwards will save in this new file. You may open this file later in order to finish an annotation. You quit dat by selecting Quit from the File menu. To attach notes to the dialog, use a comment tag or select Dialog Attrs from the Edit menu to make general comments.

The organization of tag descriptions in this manual was picked to simplify the explanation but this does not mean it is easiest to annotate in this order. You may annotate in any order you wish.



next up left folder
Next: Appendix: Outstanding Issues Up: Draft of DAMSL: Dialog Previous: Tagging Cue Words



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