Springer's Interactive Recommendation
I just discovered that Springer publishing recently incorporated an interactive recommendation tool into their website. Check this
out (You must be on the TC network). Right next to the list of related articles, there is an interactive graph that visually shows the connections of articles. You can choose different key terms or combination of key terms. The order of recommendations will change accordingly. You can play with it.
The graph is produced with D3 - a popular front end data visualization tool. The similarities or proximities are most likely calculated based on vector-space model. I think it would be very interesting to have this in our products.