Eg. Load up a retailers magazine in pdf, word file with images. Ask questions and retrieve relevant text ( easy bit) and relevant images.
eg what sizes of shoes do you have that are available in black.
Output would be say 6 seperate images with details and price below each one.