F.0.2 - Semantic Synthesis across Source Ontologies

Type: Architectural (A) Status: Stable Normativity: Normative unless explicitly marked informative

F.0.2:1 - Problem frame

Use this pattern when several sources describe the same working question through materially different concepts, relations, explanations, or boundaries, and those differences can change a named authoring decision or a contribution to one subject pattern.

Here semantic synthesis means comparing and developing claims through the concepts, relations and boundaries used by their sources. Conceptual synthesis names this semantic work; it can concern theories of action as well as other subject matters. The primary concern is one bounded semantic-synthesis question for one declared receiving use. The first useful move is to state that question and the difference that could change the next decision. Keep each source claim local through F.0.1, choose the source cut through F.1, then compare the claims here.

What goes wrong if missed. An author either leaves a crosswalk or literature narrative where a positive contribution is needed, or merges similar words before checking whether the sources describe the same entity, relation, explanation, and use. A large research package may also be demanded for a question that one bounded comparison can settle.

What this buys. The author can turn heterogeneous source ontologies into one reviewable result: a provisional synthesis claim, a contrast claim, or an unresolved-inquiry claim. When a pattern contribution is proposed, the result also states the practitioner action, first useful result, and later content decision that may accept, change, reject, or reopen it.

Not this pattern when. Cite one source directly when one source claim closes the question. Use F.1 alone when only source selection is current; F.0.1 when only local meaning is unclear; F.9 when one relation between already recovered cross-context senses is current; and the subject pattern when it already supplies the needed action and result. Use G.2 instead when the work needs a broad, refreshable SoTA Synthesis Pack@CG-Frame and its downstream Part G handoffs.

F.0.2:2 - Problem

Framework authors often receive maintained syntheses, books, papers, standards, models, and current alternatives whose ontologies are partly similar and partly incompatible. Lexical alignment can reveal candidate correspondences, but it cannot decide whether the receiving framework should adopt a shared claim. A literature summary can preserve sources, but it may never state what changes in practice.

Three failures follow:

1. a correspondence, shared label, similarity score, or framework fit is treated as permission to merge;
2. a source difference that changes action is hidden as terminology variation; or
3. missing or weak source material is reported as if the disputed domain claim were false.

Semantic synthesis must preserve source-local meanings while returning a positive next move or an honest bounded stop.

F.0.2:3 - Forces

F.0.2:4 - Solution

F.0.2:4.1 - Bounded synthesis method

1. Name the receiving question and use. State the authoring decision or subject-pattern contribution that may change. Name the practical action and first result under consideration when a pattern contribution is being proposed.
2. Recover source-local claims. For every load-bearing source, identify the claim, its EntityOfConcern, effective ReferenceScheme, edition, source role, and the context in which the claim is used. Apply F.0.1; use F.9 only when the comparison asserts a relation between exact cross-context SenseCells.
3. Choose a bounded source cut. Apply F.1. Include the sources needed for the intended use, known rival explanations, action-changing counterexamples, transfer limits, and material non-fit. State why each selected source can change the decision.
4. Compare what changes action. Compare concepts, relation participants, explanations, contexts, assumptions, source roles, counterexamples, losses, and validity limits. Treat lexical or model correspondences as proposals to inspect, not as conclusions. Use B.5.TC when the difference between theoretical accounts needs a constructed comparison: work the same question in each, locate the consequential difference and retain complementary or unresolved contributions.
5. Write one first result. Write a provisional synthesis claim when a positive shared contribution is warranted; a contrast claim when the proposed unification is not warranted for the named use; or an unresolved-inquiry claim when a named source gap prevents either substantive answer.
6. Give the result a stable local locator and name its receiving decision. Keep ordinary C.2.1 claim identity. Add a stable locator within the authoring source, such as SE-CS-03, so later text can cite the same result together with that source edition. Name the content, placement, or pattern-allocation decision that will consider it.
7. Record the receiving disposition. When the named decision is made, its decision record states whether this result is used as proposed, revised into a newly identified claim, not used for the named purpose, or reopened for further comparison. Cite the result locator and source edition. A result awaiting that decision remains a proposal; silence or a citation is not acceptance.
8. Keep the source use revisable. Identify the source claims and editions relied on, why each matters to this result, what was deliberately left local or rejected, and which change reopens the comparison.

Stop after one bounded synthesis slice when it answers the named authoring question. Enter clustering, Concept-Set, naming, UTS, or heavy harvesting only when the next question requires that additional result.

F.0.2:4.2 - Source-basis branches

F.0.2:4.3 - Three result branches

Every result recorded for later authoring, review, or decision use is an ordinary claim-bearing episteme under C.2.1. Its identity remains <claim content, EntityOfConcern, effective ReferenceScheme>. Give every such result a stable local locator in its authoring source and cite that locator together with the source edition. The locator helps readers recover the claim but does not identify it by itself.

The source gap is the reason and closure condition for an unresolved-inquiry claim, not a fourth result branch. A temporary unrecorded pause is not an F.0.2 result that later work can cite.

F.0.2:4.4 - Bounded method and broad SoTA-pack method

Use F.0.2 for one ontology-to-subject-pattern question whose smallest useful result is one of the three claims above. Use G.2 when a CG-Frame needs a broad, refreshable harvesting surface with a CorpusLedger, Claim Sheets, palette, inventories, alignment records, examples, and declared G.3-G.5 handoffs.

The methods combine in one direction. An author may use identified claims and provenance from a G.2 pack as inputs here. The author must still compare retained differences, limits, receiving use, and reopen conditions before writing an F.0.2 result. The bounded comparison does not require the rest of the pack when the question does not use it.

F.0.2:4.5 - Boundary to wording-use precision restoration

Semantic synthesis and wording restoration both pass through ontology, but they answer different questions. Semantic synthesis forms or revises a cross-source claim. Wording restoration starts from an already current subject claim and repairs a consequential wording use so that the intended entity, relation, claim kind, and admissible action can again be recovered.

A recurring wording failure may supply evidence that a subject distinction is missing or unstable. The system maintaining the affected FPF or DPF edition then uses that evidence in a named content decision. The findings do not themselves revise the synthesis claim. A DPF keeps its domain wording entries beside the domain patterns that use them; E.10.ARCH supplies the shared restoration method.

A DPF may need a reliable current domain ontology so practitioners can recognize situations, distinguish Methods and results, and use its solution moves. An author can use F.0.2 to synthesize or revise that ontology as a proposed contribution. Ontology alone is not a DPF: use E.4 and E.4.DPF to connect it to recurring problems, constructive Methods, a usable first cut, evidence practice, access, and maintenance.

F.0.2:4.6 - Semantic and methodological results

A semantic result explains what may be claimed, which distinctions matter and where source accounts agree or differ. It can include claims about actions and Methods. Methodological synthesis constructs ways of acting: selecting, adapting and connecting operations, their conditions, intermediate results and returns for a needed use. The difference concerns the question and what is constructed, not nouns versus verbs or the document in which the result is written.

Use C.39 to construct, connect, explain or change ways of obtaining the needed results. E.4.CM applies that general methodological work to a framework author’s public account of one composite Method or several independent Methods. Those constructions can use the semantic result returned here; trying a proposed way can expose a missing distinction or reopen source comparison. Neither operation requires completing a universal ontology before acting. A sufficient existing Method can be used directly.

F.0.2:5 - Archetypal Grounding

F.0.2:5.1 - Systems Engineering use and system concept

A Systems Engineering DPF author must decide what a practitioner should do when a proposed system concept is internally attractive but its connection to outside use is unclear. The author compares a proposed outside-before-inside instruction with the receiving ontology and an external systems-engineering account:

A lexical crosswalk that maps both using system and containing system to one parent system entry erases the use-versus-parthood distinction. The author can then no longer compare changing the project system’s construction with changing how an external system uses it.

The comparison concerns priority within an iterative process. Starting from an internal construction can narrow the alternatives before the outside behavior and operating conditions are known. Begin with the outside-use question. Once the information needed for a particular design question is explicit, explore construction and feasibility while refining the use account; leave a dependent construction choice open while its decisive use conditions remain unknown.

The author records the provisional synthesis branch at local locator SE-CS-USE-01 with this claim: “For one Systems Engineering concept decision, first establish the proposed outside behavior, operating conditions and system-of-interest sufficiently to compare internal constructions. Develop and test candidate constructions against that use; refine both accounts as new information arrives, and reopen the use, boundary or construction when realization evidence defeats it. State the subjects and direct relations that change the decision.” This retains the proposal’s outside-before-inside priority, reconciled with dependency-aware concurrent and iterative work.

Replay the clock choice. A team begins comparing gear arrangements for a clock. In this constructed case, asking where and how the clock will be used reveals a requirement for millisecond timing beside a rocket engine, with vibration and 30 g acceleration. The next result is an explicit timing-and-environment use hypothesis; the team compares and tests constructions against it before selecting one. Timing accuracy and vibration resistance can be investigated in parallel once their required operating conditions are stated. If feasibility evidence defeats the proposed clock use, revise that use or the system concept and compare again.

A key/lock case changes another premise: investigating the intended change from a locked to an unlocked door shifts the system-of-interest from the key to the lock, with the key treated as a subsystem in that decomposition. Recover the selected boundary before comparing mechanical, electronic or biometric realizations. The retained function stays visible while the construction alternatives change.

In this worked case, the later Systems Engineering DPF content decision cites SE-CS-USE-01 with this source edition and records used as proposed. The applied use-and-system-concept move stays in the Systems Engineering DPF; general claim, parthood and function-bearing rules stay in FPF. Reopen SE-CS-USE-01 when a source changes the linked action or boundary, a counterexample shows that the proposal cannot guide the decision, or the receiving decision changes its intended use.

F.0.2:5.2 - Organization change contrast

An organization-change author compares schools that describe organizational flows and partial organizations through different transformations and coordination structures. The sources support a narrower claim about linked transformation and coordination structures, but not one universal organization-flow kind. The author records a contrast claim, preserves the source-local decompositions, and lets a later DPF content decision choose the narrower contribution.

F.0.2:5.3 - Scientific classification inquiry

A cross-disciplinary data framework proposes one shared classification relation between a laboratory ontology and an administrative classification. The available alignment suggests candidate correspondences, but the current edition of a load-bearing clinical source is unavailable. The author records an unresolved-inquiry claim about that bounded comparison, keeps the administrative-only use available, and names inspection of the missing clinical edition as the next source action.

F.0.2:5.4 - Cheap anti-case

An engineer consults a handbook and its later edition only to recover the current pump tolerance. The later edition gives the needed claim, and no ontological difference changes the authoring decision. The engineer records direct source reliance and stops without semantic synthesis.

F.0.2:5.5 - Compare causal accounts on one sensor case

An author is deciding whether an explanation may treat inference about a hidden condition and changing an observed output as the same operation. Recover the source-local operations first. Parr and Pezzulo (2021), Introduction and The Generative Model, describes hidden-state inference together with policy-conditioned transitions and preferred outcomes. Pearl, Glymour and Jewell’s corrected primer, p.55, supplies intervention by replacement of a generating mechanism.

Use the binary load/alarm case in C.28:4.4. Under exact inference from the same stipulated joint distribution, either account obtains a high-load probability of 0.9 after an alarm and 0.1 after an ordinary low reading. In the separately specified intervention model, setting the alarm output low leaves the load mechanism unchanged, so the high-load probability is 0.5.

An active-inference action can be compared with that intervention after its state-transition and observation consequences are specified. An action that changes only the output agrees on this load prediction. An action that initiates cooling requires the later load transition and its time order in both models. Comparing which action to select also needs the objectives or preferences used in selection.

The comparison establishes this contrast: the same joint distribution supplies the stated observational inference, while agreement about an action requires a correspondence between the specified mechanisms.

The receiving explanation decision can adopt this contrast and keep the two questions explicit. Reopen the comparison when the receiving question, source account, inference approximation or action mechanism changes.

F.0.2:6 - Bias-Annotation

Scope. This pattern is limited to one bounded synthesis question in which differences between source concepts can change an FPF or DPF content decision.

This pattern favors explicit source roles and positive integration over a single authoritative vocabulary. That stance can overvalue synthesis and make every disagreement look like a new framework contribution. The cheap exits, contrast branch, unresolved-inquiry branch, and subject-pattern placement test keep the work proportional. Domain knowledge, lived practice, institutional authority, and empirical evidence remain source-local until a receiving claim states how they are used.

F.0.2:7 - Conformance Checklist

• CC-F0.2-1 The receiving question, intended use, and action-changing source difference are stated before comparison begins.
• CC-F0.2-2 Every load-bearing source claim has a recoverable EntityOfConcern, effective ReferenceScheme, edition, and source role; locators are not treated as source-use relations.
• CC-F0.2-3 The source cut covers the intended use, known rival explanations, action-changing counterexamples, transfer limits, and material non-fit by relevance rather than by a universal count or distance threshold.
• CC-F0.2-4 Similarity, correspondence, Bridge, framework fit, or G.2 pack conformance is used as input and not as synthesis truth or permission to merge.
• CC-F0.2-5 The result is one provisional synthesis claim, contrast claim, or unresolved-inquiry claim with ordinary C.2.1 identity and one stable local locator in the authoring source.
• CC-F0.2-6 The result names its receiving content, placement, or pattern-allocation decision. When that decision occurs, the decision record cites the result locator and source edition and states the disposition; an absent disposition is not acceptance.
• CC-F0.2-7 The result states relied source claims and editions, why each matters, retained differences and limits, and material reopen conditions.
• CC-F0.2-8 An unresolved-inquiry claim concerns the inquiry state, names what remains usable and the next source action, and does not turn missing evidence into a negative domain claim.
• CC-F0.2-9 Domain roles, methods, examples, evidence bases, and refresh lines stay in their subject DPF; the reusable synthesis move remains independent of those DPFs.
• CC-F0.2-10 The author stops after the smallest result that answers the receiving question and opens G.2 or later Part F work only for a named additional use.

F.0.2:8 - Common Anti-Patterns and How to Avoid Them

F.0.2:9 - Consequences

Benefits:

• framework authors can produce a useful first claim before a complete account or UTS exists;
• source differences that alter action remain visible instead of being normalized away;
• small and large synthesis investments have separate methods that can reuse the same identified source claims; and
• DPFs reuse one method while retaining their own doctrine, evidence, examples, and currentness.

Costs and trade-offs:

• authors must recover source claims and editions rather than cite files or bibliographies as wholes;
• a contrast or unresolved inquiry can be the correct result even when a positive synthesis was expected; and
• later reliance still needs the evidence-use, decision, admission, assurance, or publication claims required by that use.

F.0.2:10 - Rationale

Semantic synthesis is the middle move between preserving local meaning and publishing a unified term or subject-pattern contribution. It cannot be reduced to lexical alignment because a correspondence does not decide what the receiving framework should claim. It cannot be reduced to literature review because a source inventory does not state the practitioner action. It should not require a full G.2 pack because many authoring decisions need only one bounded comparison.

The three result branches keep the method constructive without forcing agreement. A provisional synthesis opens a positive contribution, a contrast preserves a decision-relevant difference, and an unresolved inquiry turns a real source limitation into a defined next action. Ordinary C.2.1 claim identity keeps those results separately revisable.

F.0.2:11 - SoTA-Echoing

For the authoring question in :5.1, a correspondence-only answer can identify similar terms but leaves the outside-before-inside contribution undecided. A full G.2 pack adds a reusable corpus, claim sheets, inventories, alignments and downstream handoffs; these become worth their cost when later questions need that broader result. The bounded method instead retains source-local claims, compares the decision-changing difference and produces one result for the named decision.

Jaakkola’s question and source-role discipline, Okoli’s extraction and iterative synthesis, and Carroll–Brunton’s treatment of non-fit inform steps 1–5 of that bounded route. These contributions organize the comparison; the particular working case selects what can be integrated and what must remain different. The accepted trade-off is limited coverage for a smaller first result. Reopen the source cut when a known rival or counterexample can change that result, and open G.2 when the receiving work needs its broader reusable outputs.

F.0.2:12 - Relations

• Builds on: F.0.1 for source-local meaning and F.1 for a relevance-based source cut.
• Uses when current: F.9 for an asserted relation between exact cross-context SenseCells; C.2.1 for every recorded result claim; A.2.4 and the applicable direct relation pattern for a later evidence or decision use.
• Coordinates with: F.2-F.8, F.14, F.17, and F.18 when the next authoring question requires harvesting, clustering, Concept-Set, naming, or UTS work.
• Coordinates with: G.2 as the optional broad harvesting method. Identified claims and provenance from a G.2 pack may supply inputs here; the pack does not replace the receiving comparison.
• Coordinates with: E.4.DPF for DPF entry and result placement, E.10.ARCH for DPF-local wording entries under the shared restoration method, and G.11 for source-currentness and refresh work.
• Coordinates with: C.39 for general Method construction, connected explanation and change, and E.4.CM for their application to framework authoring. Their methodological results and this semantic comparison remain distinct and can revise each other.
• Boundary: Domain claims produced through this method remain in the subject FPF pattern or named DPF selected by the later content decision.

F.0.2:End