A.6.9:11 - SoTA-Echoing
(informative; post-2015 alignment)
| SoTA practice | Primary source | What A.6.9 echoes | What A.6.9 adds | Stance |
|---|---|---|---|---|
| Correspondences between viewpoints | ISO/IEC/IEEE 42010:2022 | Correspondence is not identity and retains intent and constraints. | Separates the direct semantic relation from each proposed use and actual publication or view object. | Adopt + specialise |
| Declarative validation shapes | W3C SHACL (2017) | Make implicit conditions testable. | Uses a profile for relation truth, a claim for bounded-use suitability, and a card only for packaging. | Adapt |
| Scored entity alignment with error analysis | BootEA (Sun et al., 2018) | Alignment evidence is graded and fallible. | Keeps scores and counterexamples as evidence rather than relation identity or a use licence. | Adapt |
| Textual entity matching | BERT-INT (Tang et al., 2020); Ditto (Li et al., 2021) | Matchers yield conditional, error-prone correspondences. | Requires exact endpoint readings, a falsifiable Bridge predicate, and a separate action-specific claim. | Adopt conceptually |
| Heterogeneous schema matching | SMAT (Zhang et al., 2021) | “Match” covers several relation types. | Distinguishes relation kind, relation orientation, proposed-use direction, rule, and tolerance. | Adapt |
| Human-in-the-loop matching | Mudgal et al. (SIGMOD 2018) | Scores require abstention and curated error cases. | Uses the exact A.10 evidence or B.3 assurance predicates and preserves explicit negative or blocked outcomes. | Adapt |