Entity resolution, also known as name resolution, refers to resolving a reference. The usual language continues with something to the effect “…real world object, person, etc.” I omit the usual language because a reference can be to anything.
Do you disagree? Just curious.
I have more comments on “resolution” but omit them here to reach the necessity for identification.
I say the “necessity for identification” as a prerequisite before semantics can be assigned to any identifier. Of whatever nature. Could be text, photograph, digital image, URI, etc.
I say that because identification (or recognition, a closely related task) may have different requirements (processing and otherwise) than the assignment of semantics, or the assignment of other properties.
For example, with legacy text, written on read-only media, if my only concern at the first step in the process is identification, I can create a list of terms that represent the terms I wish to have recognized. (Think of the TLG for example.) The semantics that I or anyone else wishes to associate with those terms becomes an entirely separate matter. Whatever means or system of semantics is used.
That only becomes possible if the notion of assigning semantics is separated from the task of identification. And separated from the organization of what has been identified and assigned semantics into a system (think ontology).
(You might want to read what was possible twenty-two (22) years ago with transient nodes and edges before responding to this post. I would extend that type of mechanisms to recognition, assignments of semantics and other properties. Having a fixed identification, assignment of semantics and other properties being only one choice among many.)