There seem to be a number of high-level terms in the evidence ontology that have children defining whether the evidence was obtained from automatic or manual assertion methods, e.g. 'combinatorial analysis' (ECO:000212) has the children 'combinatorial analysis used in automatic annotation' and 'combinatorial analysis used in manual assertion'.
When a database has imported a set of annotations where it is unable to identify whether an assertion has been made manually or automatically - could it use such a top-level term, so as not to provide any indication as to the manual or automatic nature of the method?
N.B. such a term would not be intended for GO annotation.