The difference is slight, but consistent. Transforming a representation into another one involves taking the same data and expressing it in another format - for instance, a hexadecimal into a decimal number, or mixed-case strings to an all-lower string. Typically, you need nothing but a fixed set of rules to carry out such a transformation.
Resolving a name typically involves querying some repository that associates names with information that cannot be predicted without the repository.
google.com may resolve to a specific I.P. address (several, actually), but that is an accident of the I.P. assignment history - it might just as well have been a totally different number, and there is no way of making this translation that doesn't somehow involve storing the entire registry, or querying it online. Similarly, resolving packages dependencies usually requires installing the missig packages - it isn't enough to know the names of the missing components, you have to actually acquire their contents, and the contents are not predictable from the name alone ("QMail" sounds like a mail program, but you can't predict its exact properties just from the name without installing it, since the name transports far less information than the compiled program).
Therefore you shouldn't use the term "resolving" for something that is a mere context-free data transformation. Reserve it for situations where a substantial amount of information is symbolized by each name, so much that the only reasonable arrangement is to have a central registry for it.