K. Prashant worked on this.

Here is an abstract:

In this project we have developed a method to  annotate documents
generated electronically.  Our method finds most common use in previewers for
documents generated using LaTeX.  Document preparation systems like LaTeX
generally take text along with directives to a compiler as input and produce
device independent files as output.  In order to view these device
independent files on a display, previewers are used.  One such previewer for
LaTeX in the X--Windows environment is  xdvi.
 
When a user, who is not necessarily the author previews a document, she may
find the need to add comments about some items in the document.  Current
previewers do not support this feature.  This deficiency is sought to be
corrected.
 
In our scheme, to annotate an item the user selects a portion of the text
rendered as a raster image.  Since these annotations must be redisplayed when
the revised document is previewed again (by the author for instance), the
program needs to know what portion of the document this item corresponds to.
Our method uses document analysis and character recognition techniques to
accomplish this.