Although these guidelines deal purely with syntactic annotation of corpora, the problem of the boundary between syntax and semantics necessarily arises. There are cases where a phenomenon clearly falls within a semantic analysis rather than a syntactic one, but there are also cases where the boundary is slightly more fuzzy. For example, it is common practice within most approaches to grammar to subcategorise adverbial clauses as adverbials of e.g. `place', `manner' or `time'. These categorisations are primarily semantic -- i.e. the choice of category is made dependent largely on semantic information -- in this case the meanings of the words in the adverbial clause. However, these semantic categories involve syntactic properties, for example in English whether an adverbial is of `place' or `time' will affect its position in the sentence.
Although a semantic analysis is beyond the scope of the guidelines outlined here, mention will be made of some more semantically-based distinctions. The boundary between syntax and semantics is itself much disputed, and in terms of NLP, it has been shown through work on POS tagging that often semantic information may be an aid to automatic class identification.