A brief explanation of the annotation format:

-------------------------------------------------------------------
    article [40332]
      right-click the folder and then click properties
         c:0 u:2 p:-1
         - right click (65618-0-l-5-85) [Item] Folder
         c:5 u:6 p:-1
         - left click (1245346-0-m-30739-12) [Menu] Properties
      on the general tab click advanced
         c:0 u:2 p:-1
         ~ left click (327956-0-c-0-102) [Tab] General
         c:4 u:5 p:-1
         - left click (262280-524588-w--1-83) [Button] Advanced...
-------------------------------------------------------------------

In the above example snippet:

1. lines starting with 'article [' contain the document id.  The document
id corresponds to the id in the Windows 2000 dataset.

2. lines start with 'c:' identify the selected word indices, where the
numbers following c,u,p are the indices of the command word, object
word, and text parameter respectively. p:-1 indicates a null text parameter.

3. lines starting with '-' or '~' contain the command, object id, object
type and object name in the following format: "- <command> (<object id>)
[<object type>] <object name>".  If the line starts with a '~', the
command is optional.  In the above snippet, clicking on the general tab
is optional because the folder properties dialog opens with the general
tab selected.  At the same time, following the instructions and clicking
on the tab is not incorrect.  

Note: The "object id" is a string constructed by the OS agent for each
interface object, and is guaranteed to be unique within a given state.
This id is constructed based on the HWND handles of the UI object's 
window, it's parent in the UI hierarchy, and so on.  These ids are
not guaranteed to be unique and constant over time (i.e. multiple states)
for the same UI object.  I.e. the same dialog box opened twice in 
sequence can have different object ids.

4. The lines with no specific marker at the beginning are the sentences 
from the document.  So all word index lines and command lines refer to the
previous sentence line.