In PLSR, the "beta" values are the regression coefficients that can be multiplied by your X data to give you your predicted Y data. You can plot the beta values associated with each x variable to give an indication of how strongly each x variable contributes to each latent variable.
For example, say you are modelling tree age (y) based on height(x1) and trunk width(x2), and density(x3) using two latent variables. If you have centered your data before PLSR, then you would get beta values along the lines of B=[b11,b12;b21,b22]. Your predicted y vector would be y=B*X. If you plot the two points (b12,b22) and (b21,b22), then the distance of each of these two points along each axis will show how strongly each variable contributes to the component associated with that axis (e.g. component 1 = x-axis; component 2 = y-axis).
The correlation coefficient (r) would be something you use afterwards, to compare how well your model explains the actual variation in the tree age (correlation between actual age and predicted age). Or you could look at the relationships between the individual X variables. The partial correlation coefficient between two variables measures what the correlation would be if all other influencing variables were held constant at their means. For example, you could find the partial correlation coefficient between x1 and x2 while controlling for the effect of x3.