ACE Annotation Task Tutorial

The ACE annotation task is also known as the RDC annotation task.

Opening files for annotation

There are three ways to begin annotating a file:

  1. File->New will create a new annotation file for a raw data file
  2. File->Open will open an AIF file previously created by Callisto.
  3. File->Import will use a Task specific mechanism to convert from some other data type to the task specified

The RDC Task defines a single importer for the ACE Pilot Format (APF).

All data annotated by this version of Callisto are saved in AIF format. Tasks may also define Exporters, which would convert to an external format of the Tasks choosing.

Note that when loading an APF file, the "signal" file (or "source document") annotated by the APF stand off annotations will be a separate file. This source document file annotated by the APF is indicated within the APF itself and must (obviously) be present in order to perform annotation. The APF2AIF conversion process that happens automatically when loading an APF file is able to find the name of this source document file. Since the source document file is specified without any directory prefixes, it is assumed and required that the source document is located within the same directory as the APF file being opened. The APF2AIF conversion process then creates a raw text file (called ${source-document}.txt, where ${source-document} is the name of the source document annotated by the APF, *not* the APF file itself).

In order to load annotations created by an earlier session with Callisto, one needs to load the AIF file associated with those earlier annotations. Thus, even if one specified a raw text file "foo.txt" as input to an earlier Callisto annotation session, all the annotations associated with that session will be saved in an AIF file (for example, "foo.txt.aif"). To see and update these earlier annotations one must open the AIF file ("foo.txt.aif"). Since AIF, like APF, is a stand off XML annotation format, there are mechanisms within the AIF file that refer to the original raw text file (referred to in the ATLAS annotation model as the "signal"), so Callisto will be able to retrieve this original text file in order to properly display it along with the annotations on that file. (See the bug/limitations notes below on moving or transporting AIF files.)

To make it easy to revert to earlier versions of a set of annotations, Callisto generates backup copies of the annotations files that are saved separately from the primary named annotation file. The format of these backup files is *.aif.${integer}.xml, where ${integer} takes on ever increasing values as new versions are saved to disk. The user can select in the preferences window how many previous versions are to be maintained at any given time.

Here are some examples of entering files into Callisto, using sample ACE (RDC) data found in the ${CALLISTO}/data/ACE/ directory. Entering (and saving) files involve the use of "File" drop-down menu, which is in the upper left-hand corner of the Callisto window (circled in dashed red in the image below):

[callisto window image]
  • Enter a "New" raw file to be annotated (like a *.txt file in the samples):
    1. Mouse on: "File" -> "New"

      A pop-up window will appear.
    2. Browse to get to the new file to enter
    3. Select "RDC Task" for "Task"
    4. Uncheck the "Parse as SGML/XML" box
    5. Leave the file encoding at UTF-8
    6. Press "Annotate"
  • "Open" an existing AIF file (like an *.aif.xml file in the samples):
    1. Mouse on: "File" -> "Open"

      A pop-up window will appear.
    2. Browse to get to the file to open
    3. Press "Open"
  • "Import" a APF format file (like an *.apf.xml file in the samples):
    1. Mouse on: "File" -> "Import"

      A pop-up window will appear.
    2. Browse to get to the file to import
    3. Select "RDC Task" for "Available Importers"

      "APF" will be a forced choice.
    4. Leave the file encoding at UTF-8
    5. Press "Import"

Saving annotations

To save your annotation, you must choose a file name which will be recognized as AIF, having an extension as listed above (*.aif, *.aif.xml, *.aif.*.xml). Once saved, do not move the signal file. Callisto uses standoff annotation, and enforces a 'no change to the signal' policy that includes the signal's location. If the signal file is moved in any way, (even if the annotation file moves with it) you will no longer be able to reload the annotations (see Known Bugs and Limitations #1 below, for more information). This restriction can be surmounted by carefully editing the AIF file itself. If you need to do this, please contact MITRE.

To save out the annotations (using the AIF format):

  1. Mouse on: "File" -> "Save As"

    A pop-up window will appear.
  2. Browse to where you want the saved file to go
  3. Change the file name to save out to what you would like
  4. Press "Save"

The Callisto Interface for ACE

Upon opening a file for ACE annotation in Callisto, one is presented with a main text screen in which the raw text of the original file is displayed. At the bottom of the main window are a number of distinct "sheets" that can be displayed in the same space by selecting the appropriate "tab" indicating their contents: Mention, Entities, Relations. These tables are used to display the various kinds of ACE annotations. A useful layout is to "tear off" the Entity and Relations tables to enable them to be placed in a separate location on the user's desktop. This can be done by clicking on the icon (looks like two overlapping rectangles) immediately to the left of the tab labels. Another useful "default" layout selection is to select (check the box) of the "Text Palette" option underneath the "Edit" pull down menu.

Text in the main text viewing window can be selected by swiping the mouse while depressing the left button. The default behavior is to have the tool itself automatically select words at what it deems are word boundaries. If this default behavior interferes with sub-word selections, the user can change this preference from the "Selection Mode" option underneath the "Edit" pull down menu (changing it to Character mode). Selecting text does not itself create a mention annotation. A mention annotation is created when text has been selected and then either (a) the right button is selected and an appropriate action is selected from the pop-up menu, (b) the desired action is selected directly from the "Text Palette" widget, or (c) the user presses the middle button, in which case the default action associated with the "Text Palette" widget (indicated by the black circle button in the right hand column) is taken. It will often be useful to have the default action be "New Mention", which will enable the annotator to only rarely be forced to use either the text-centered pop-up menu of actions or the text palette actions.

Mention annotations can be selected by clicking on them (by default the most recently created mention is the current selection). When selected, a small line on top and below the annotated text appears to indicate that it is the selection mention. If a given bit of text has multiple mention annotations, the user can place the mouse directly over the multiply-annotated text and hit tab to "cycle" through the various co-extensive annotations.

Creating a mention creates two phrases at once that are linked together: the "full" extent of the mention, and the "head" of the mention. A new entry in the Mentions table is created in which the various attributes of the mention are displayed explicitly. By default the head and full extent of a mention are co-extensive. To change the extent of either the head (the usual case when a modification is necessary) or the full extent, one swipes the desired text to create a new selection, then either (a) clicks the right button and selects the "Modify Head Extent" entry of the pop-up menu, (b) selects the same entry from the free-standing "Text Palette," or (c) clicks the middle button if it so happens that the user has selected "Modify Head Extent" as the default action in the Text Palette window (indicated by the black circle to its left). Attributes of mentions besides their textual extent can be modified directly from within the mentions table.

To create an EDT entity, one selects a row in the mention table or selects a mention annotation from within the text window pane, and then clicks right to get the pop-up window. This will present "Add to Entity" and "Create new Entity" as two options. If one selects "Add to Entity", the tool expects the user to immediately follow that selection with a left mouse click on the row in the entity table that indicates the entity to which the mention should be attached. A similar action, but with a different action selected, allows one to add a selected mention to a relation (either as an Arg1 or Arg2 filler. The other attributes of entities and relations are set just as in the mentions table --- these values are all drop-menus that are set by clicking on the cell and selecting from the pop-up menu.

Sample ACE data

Some sample ACE data is provided in the ${CALLISTO}/data/ACE/ directory. There are four types of files:

  1. *.sgm.tmx.rdc.apf.xml: An APF file. Can input into Callisto via "Import".
  2. *.sgm: An sgml file. In these samples, they are the "signal" files that the *.sgm.tmx.rdc.apf.xml files point to.
  3. *.sgm.tmx.rdc.apf.xml.txt.aif.xml: An AIF file. What Callisto saves out. Can input into Callisto via "Open".

    Note: some versions of Callisto save out in a slightly different format. But these sample files are still readable by Callisto.
  4. *.sgm.tmx.rdc.apf.xml.txt: A plain text file. Can input into Callisto via "New".

    In these samples, they are also the "signal" files that the *.sgm.tmx.rdc.apf.xml.txt.aif.xml files point to.