Semantic Agreement. Semantic agreement is determined by the meaning of the controller, i.e. is semantically justified. It is consistent with the properties of the referent. Semantic agreement is traditionally contrasted with syntactic agreement, which is consistent with the formal properties of the controller. As the terms syntactic and semantic agreement can be said to represent opposing sides of a spectrum, rather than some particular kinds of agreement in isolation, they are most meaningful when there is a choice between the two (Xxxxxxx 2006:155). Consider [13]: [13] Наш бухгалтер приш-л-а получить транш 1PL.POSS.M accountant(M/F) come-PST-F get tranche 8 This variation is also reported with other examples in Xxxxxxx 1983:52-55, 2006:232-233 as evidence for the Predicate Hierarchy. от американского фонда XXX и узна-л-а, что перечисления from American fund XXX CONJ learn-PST-F that transfers невозможны. impossible ‘Our accountant came to get the tranche from the XXX American fund and learnt that the transfers are impossible’ (xxxxx://xxxxx.xx/review/view/16608/) [13] exhibits both syntactic and semantic agreements. The possessive pronoun наш ‘our’ in attributive position agrees with the controller бухгалтер ‘accountant’ according to its formal properties: its morphological form is masculine. The predicate verbs пришла ‘came’ and узнала ‘learnt’, however, show feminine agreements, based on the meaning of the controller: the referent is a female accountant. As shown in [13], both syntactic and semantic agreements can be found in the same domain of a clause. Notably, syntactic agreement occurs in attributive position, whereas two instances of semantic agreement in predicative positions. This particular distribution is what is called an Agreement Hierarchy effect: the greater the syntactic distance between the target and the controller, the more likely is the semantic agreement (Xxxxxxx 1979:223). The Agreement Hierarchy will be introduced in the next section.
Semantic Agreement. The key feature of the proposed semantic agreement approach is the agreement unit, which is used to express semantic mappings between the concepts exported by peers and the element of the shared ontology. Each agreement unit defines a semantic interpretation or view of one or more ontology concept, thereby adapting the semantic of global shared concepts to the local view and constraints of a peer. An agreement unit defines 1:1, 1:N, M:1 and M:N SP PP SP PP mappings between ({COm }, {ESn }) where COm is an ontology and ESn is an export
Semantic Agreement. An important difference between our model and those of [7, 9, 31] is that in [7, 9, 31] epochs are (uniquely) identified by messages creating them. This is problematic for saCGKA, because different receivers transition to a given epoch using different messages. Cru- cially, this means an injected message cannot be used to identify the injected epoch into which its receiver transitions. We deal with this in a clean way by allowing the simulator to identify epochs. That is, epoch identifiers are arbitrary as long as consistency, authenticity and confidentiality hold. The work [31], which proposes a new CGKA where, similar to SAIK, receivers get personalized packets, encountered the same problem with the existing models [7, 9]. In their new model, filtered CGKA (fCGKA), an epoch is identified by the sequence of packet headers leading to it. The header is a part of the uploaded packet that is downloaded by all receivers. A protocol can be secure according to the fCGKA model only if the header it defines has the properties of a cryptographic commitment to the semantics of the packet. saCGKA generalizes fCGKA (and [7, 9]) and provides additional flexibility. For instance, it enables CGKAs which, like SAIK, assume PKE with the weaker RCCA security, while fCGKA still requires the stronger notion of CCA. We believe that in the future more CGKA protocols will take advantage of saCGKA’s flexibility. For example, one may consider using a different packet-authenticator for each receiver with the goal of providing some level of unlinkability – an adversary seeing only packets downloaded by participants cannot tell if they are in the same epoch (or group) or not.
