Structure Transfer Calculus

• Structure Transfer Calculus is an inference-based method that transforms representations across diverse representational systems while preserving semantic equivalence.
• It employs construction spaces, multi-space systems, and pattern graphs to model abstract representations in various formal and informal notations.
• The calculus utilizes schemas as inference rules to guide both complete and partial representation transformations, enabling applications such as diagram generation and formal analysis.

Searching arXiv for the specified and closely related papers to ground the article in current metadata. arXiv search query: (Raggi et al., 3 Sep 2025) Structure Transfer an Inference-Based Calculus for the Transformation of Representations Structure transfer calculus is an inference-based calculus for transforming representations across diverse representational systems (RSs). In its defining formulation, a source representation from a source RS is used to generate a target representation in a target RS such that the source and target satisfy a specified relation, including semantic equivalence. The calculus is explicitly representational-system agnostic: it is formalised in Representational Systems Theory (RST), uses construction spaces to model RSs of diverse kinds, and relies on schemas that encode preservation of information across relations between systems (Raggi et al., 3 Sep 2025).

1. Conceptual scope

The central problem addressed by structure transfer calculus is the representation transformation problem: given a representation in one system, construct a representation in another system while preserving a desired relation. The calculus is presented as a general method for representation transformation and choice, rather than as a translation procedure tied to one formal language or one semantic domain (Raggi et al., 3 Sep 2025).

Its generality depends on two commitments. First, representations are treated abstractly through construction spaces, so the underlying formalism can model formal languages, geometric figures and diagrams, as well as informal notations. Second, the target of transformation is constrained relationally rather than only syntactically: the generated target must satisfy a specified relation to the source, and schemas encode the knowledge needed to guarantee that relation (Raggi et al., 3 Sep 2025).

A recurrent misconception is to treat structure transfer as a direct encoding map between two notations. The formalism is broader than that. The relation between source and target need not be a fixed translation function; schemas can express preservation of information across relations between any pair of RSs, and the transfer process can proceed by abductively constructing a target representation whose structure is sufficient for the chosen relation (Raggi et al., 3 Sep 2025).

2. Formal substrate: construction spaces, multi-space systems, and patterns

The basic semantic and syntactic unit is the construction space

$\mathcal{C} = (T, C, G),$

where $T$ is a type system, $C$ is a constructor specification, and $G$ is a structure graph encoding all possible constructions in the system (Raggi et al., 3 Sep 2025).

The type system is given as $T = (Ty, \leq)$, with types and a subtype relation. The constructor specification is written $C = (Co, sig)$, with signatures of the form $sig(c) = ([\tau_1,\ldots,\tau_n], \tau)$. The structure graph $G$ functions as the field of possible representations admitted by the system (Raggi et al., 3 Sep 2025).

To relate multiple representational systems, the calculus uses a multi-space system

$M = (\mathcal{C}_1, \ldots, \mathcal{C}_n, G),$

where each $\mathcal{C}_i$ is a construction space and the final component is a meta-space encoding relations or properties involving tokens from the component spaces (Raggi et al., 3 Sep 2025). This enables cross-system constraints such as depiction, equivalence, or other relational conditions to be represented alongside the structures they relate.

Pattern graphs provide the operative templates for reasoning. A pattern graph is a structure graph that respects the type and construction constraints but may be incomplete or underspecified. It therefore functions as a reusable template for matching and constructing structures within or across construction spaces (Raggi et al., 3 Sep 2025). Pattern manipulation is mediated by morphisms, including instantiation, reification, and loosening. Reification makes a pattern more concrete by specialising types or adding structure; loosening makes it more general by replacing types with supertypes or removing structure and context (Raggi et al., 3 Sep 2025).

Notion	Formal role	Representative form
Construction space	Models one RS	$T$0
Multi-space system	Relates several RSs via a meta-space	$T$1
Pattern graph	Template for matching or construction	underspecified structure graph

3. Schemas and the inferential mechanism

Schemas are the basic units of knowledge in the calculus. They encode invariance information as generalised inference rules, and their purpose is to state how structural or semantic relations are preserved within or across representational systems (Raggi et al., 3 Sep 2025).

Formally, a schema is given as a sequent. In the presentation summarised for the paper, the context consists of pattern graphs for the component construction spaces and the consequent is a pattern in the meta-space. The intended condition is that, for any instantiatable specialisation of the context patterns, there exists a corresponding instantiatable specialisation of the goal, preserving the desired relationships (Raggi et al., 3 Sep 2025).

A transfer schema is a schema equipped with a set of target indices $T$2. These indices designate the construction spaces in which structure may be transferred. Operationally, transfer schema application permits reification of the target-side context patterns while the source-side patterns remain fixed. This is the mechanism by which a target representation can be built “backwards” from relational requirements (Raggi et al., 3 Sep 2025).

The calculus supports both forward and backward schema application. Forward application adds inferable consequences to the assumptions. Backward application reduces a goal to subgoals determined by the schema. In both directions, refinement maps, reification functions, and loosening maps mediate how patterns are grown, specialised, or abstracted during inference (Raggi et al., 3 Sep 2025).

This inferential design implies that transformation is knowledge-driven rather than purely syntactic. The target structure is not merely rewritten from the source; it is assembled through schema-guided satisfaction of relational constraints. A plausible implication is that representation change is treated as a constrained construction problem rather than as compilation.

4. Transfer procedure, validity, and search

The transfer procedure starts from a sequent that specifies the source, a minimally specified or empty target, and the desired relation in the meta-space. It then iteratively applies $T$3-transfer schemas, reifying the target context so that more schemas become applicable. When a valid sequent is reached, the target pattern graph encodes a structure in the target construction space that, if instantiated, stands in the desired relation to the source (Raggi et al., 3 Sep 2025).

The key metatheoretic notion is validity of a sequent. In the summary associated with the paper, a sequent is valid if it can be obtained by a possibly empty sequence of schema applications from base schemas. The accompanying informal theorem states that if a valid sequent

$T$4

is obtained by transfer schema applications allowing reification of the target, then for any instantiation of the source there is an instantiation of the entire pattern that satisfies the desired relation (Raggi et al., 3 Sep 2025).

The procedure is explicitly non-deterministic. There may be many possible transfer-schema applications, and heuristics, domain knowledge, or search strategies may be used to guide the process (Raggi et al., 3 Sep 2025). This means that the calculus specifies a principled inferential space for transformation, not a unique normal-form derivation.

A second misconception is that the calculus guarantees only complete transformations. The formalism also allows partial or unverified transformations when schemas or knowledge bases are incomplete. This follows from the stated support for partial knowledge and for partial transformations in the presence of incomplete schema information (Raggi et al., 3 Sep 2025).

5. Canonical example and range of applications

A representative example is the transformation of the set-algebra statement

$T$5

into an Euler diagram while ensuring that the diagram semantically depicts the original logical statement (Raggi et al., 3 Sep 2025).

In this example, the source pattern graph encodes the set-algebra expression, while the meta-space expresses the relation “depict.” Transfer schemas capture facts such as depicting a conjunction by depicting both conjuncts, and depicting a subset by constructing regions with the appropriate containment relation. By backward application of these schemas, repeated reification of the target diagram adds regions, labels, and containment constraints until the resulting target pattern graph encodes the structure required for semantic depiction (Raggi et al., 3 Sep 2025).

The stated application range is broad. The calculus is presented as applicable to automated diagram generation from logical or formal input, mapping between formal and informal notations in scientific software and theorem provers, analogical reasoning, creative and adaptive problem solving, and meta-level formal methods that generalise term rewriting, simulation, abstraction, and refinement transformations (Raggi et al., 3 Sep 2025).

Because the underlying formalism abstracts over the notion of representation, the same machinery can in principle operate on sentential, diagrammatic, hybrid, and informal systems. This suggests that the main contribution is not a domain-specific representation converter but a general method for justified re-representation across heterogeneous symbolic media.

6. Related uses of “transfer” and “calculus” in adjacent literature

The phrase “structure transfer calculus” is domain-sensitive, and adjacent literatures use closely related language for substantially different objects.

In homological algebra, “Transferring algebra structures on complexes” studies transfer of dg algebra structures along deformation retracts and perturbed homotopy equivalences. Its basic transferred product has the form

$T$6

and, after perturbation,

$T$7

There the phrase refers to explicit homological machinery for descent of algebraic structure, controlled by generalized or scaled Leibniz conditions on the homotopy, rather than to RS-agnostic representation transformation (Miller et al., 2020).

In properadic homotopy theory, “Properadic homotopical calculus” generalises operadic calculus to properads, introducing monoid $T$8-comodules, properadic $T$9-morphisms, and a homotopy transfer theorem with graph-based transfer formulas. Here transfer concerns homotopy gebras over a properad and explicit formulas indexed by connected graphs with multiple inputs and outputs, not transformation between heterogeneous representational systems (Hoffbeck et al., 2019).

In concurrency theory, “Classical Transitions” describes a calculus in which hypersequents register parallelism in typing judgements, thereby enabling a transfer of structure between classical linear logic and $C$0-calculus-style processes. The technical focus is typing, labelled transition semantics, subject reduction, and progress, rather than construction-space-based transformation (Montesi et al., 2018).

A further possible confusion arises with the calculus of structures. “Computational Aspects of the Calculus of Structure” studies deep inference in CoS and proposes proof-search restriction for the FBV subsystem via the incoherence number of substructures. In that setting, “structure” refers to logical structures manipulated by inference rules such as $C$1 and switch, and the central issue is taming non-determinism in proof search, not representation transformation across RSs (Alvim, 2013).

Taken together, these neighboring literatures show that “transfer” and “calculus” recur in several technical traditions. The distinctive feature of structure transfer in the sense of (Raggi et al., 3 Sep 2025) is that it treats representation change itself as an inference problem over abstract construction spaces and schemas, with correctness expressed relationally rather than solely algebraically, operationally, or proof-theoretically.