In Adobe, it is a bit difficult, as you would need a really particular kind of mud…
But I guess you are looking at Adobe Acrobat to do this. And here it is possible. It will require a bit of Acrobat JavaScript programming, but can be done. Roughly, what you would do is this:
Either create a first annotation, and then select it, or enter the keyword in a dialog.
Search the document for the occurrence of the keyword and read out its coordinates on the according page.
Create an annotation (such as highlight) at the place given by the coordinates from step 2.
And that would be it.
In order to implement, you would get the Acrobat JavaScript documentation, which is part of the Acrobat SDK documentation, freely downloadable from the developer section of the Adobe website.
If the performance of Acrobat JavaScript is not high enough, you'd look around for a developer who would build a custom production plug-in for Acrobat. However, note that this would not work in Reader, whereas the JavaScript solution would.