According to many information scientists, it is useful to think of information as part of this sequence: Data -> Information -> Knowledge, where the following definitions make sense:
Data are sense stimuli that have been attended to (Noise -> Attention -> Data).
Information is sensible structured data (Data -> Structure -> Information).
Knowledge is usable information that has been interpreted through a social-cultural process (Information -> Interpretation -> Knowledge).
There are many sense stimuli around me, but my attention turns to two moving bodies below. At that point I have data - two moving bodies.
Right after that, I ascribe meaning to these data: I structure them by classifying them into pre-existing categories - here's a man, here's a woman, here's a street. Now, I have information/meaning: I see a man and a woman walking down the street.
Can I do anything with this information? Do I care? Probably not. If I do care, I can
interpret this information in the light of various other information pieces in order to create some kind of contextual picture which other people find sensible and usable.
Let's say we're waiting for a couple to visit us. I focus on the man and the woman, but can't see their faces. I call my partner from the other room, and ask her - Can you see there? Are these John and Wendy? And she says - "yes, they are". I look again and I say - "yeah, you're right, let's call them and help them find our apartment". Now we've created a usable piece of knowledge through a social process.
So the sequence was:
Noise (various sense data around)
-> Attention -> Data (two moving bodies) -
> Structure -> Information (Man and woman walking down the street)
-> Socio-Cultural Interpretation -> Knowledge (John and Wendy coming to visit us and might need help in finding our apartment).
You can also check out a nice post by Khen Ofek at his blog 'Cloud Philosophy'.