{"schema_version":1,"research_id":"eoa_inverse_innovation_exp05_external_evaluation_20260803","source_assessment_id":"invariant_mode_decomposition_design__literature_literary_theory:P2:v0","cell_id":"invariant_mode_decomposition_design__literature_literary_theory","search_queries":["site:mla.org Committee on Scholarly Editions guidelines editions verification apparatus cross references","TEI Guidelines critical apparatus link notes cross-reference dependency scholarly edition","critical edition errors apparatus inconsistencies editorial workflow source verification","scholarly edition claim dependency graph provenance humanities research","MLA Committee Scholarly Editions Guidelines accuracy verification annotation apparatus official","site:scholarlyediting.org critical edition workflow errors correction apparatus notes dependencies","digital scholarly edition dependency graph apparatus editorial notes correction network","critical edition quality control proofreading verification annotation errors scholarly editions","digital scholarly edition knowledge graph claims evidence dependency critical apparatus ontology verification","argumentation graph scholarly editing textual criticism claims dependencies digital humanities","scholarly editions graph provenance contradiction resolution humanities primary research","critical apparatus knowledge graph dependency claims corrections","NetworkX eigenvector centrality documentation directed graph left eigenvector strongly connected convergence","NetworkX pagerank documentation directed graph damping official","SciPy eigs documentation sparse matrix eigenvalues official","spectral centrality instability perturbation nearly degenerate eigenvalues network paper","W3C PROV-O wasDerivedFrom wasInfluencedBy official recommendation provenance ontology","W3C Web Annotation Data Model motivation linking body target official recommendation","copyright unpublished manuscripts scholarly edition permissions archival sources official guidance","MLA scholarly edition copyright permissions digital edition guidelines","Guide to Documentary Editing third edition publication date Kline Perdue University Virginia Press","site:gde.upress.virginia.edu \"Guide to Documentary Editing\" \"2008\""],"sources":[{"source_id":"S1","title":"Guiding Questions for Vetters of Scholarly Editions","publisher":"Modern Language Association of America, Committee on Scholarly Editions","url":"https://apps.mla.org/pdf/guidelinesform.pdf","source_class":"OFFICIAL_GUIDANCE","publication_date":"2003-09-15","accessed_at":"2026-08-03","claims_supported":["An identifiable scholarly-edition authorizer expects independent transcription checks, sampled accuracy testing, accurate and consistent apparatus, and documented emendations.","The checklist asks electronic editions to maintain correction files, use community encoding guidance, validate schemas, document databases, and enforce referential integrity.","These requirements establish institutional demand for traceability and correction control, but do not express demand for spectral prioritization."]},{"source_id":"S2","title":"A Guide to Documentary Editing, Chapter 6: From Transcriptions to Editorial Texts","publisher":"University of Virginia Press, in association with the Association for Documentary Editing","url":"https://gde.upress.virginia.edu/06-gde.html","source_class":"AUTHORITATIVE_SECONDARY","publication_date":"2008-10","accessed_at":"2026-08-03","claims_supported":["Editorial transcriptions require repeated review, and informational annotation and quotations require verification.","Verification of note elements should be performed by someone other than the original annotator.","Projects should maintain a record of corrections identifying what changed, who changed it, and when.","Time and staff constraints can prevent the amount of independent checking editors would prefer, supporting the relevance of bounded prioritization."]},{"source_id":"S3","title":"TEI P5 Guidelines, Chapter 13: Critical Apparatus","publisher":"Text Encoding Initiative Consortium","url":"https://guidelines.tei-c.de/en/html/TC.html","source_class":"STANDARD","publication_date":"2026-02-18","accessed_at":"2026-08-03","claims_supported":["Critical editions contain highly structured apparatus entries connecting readings, witnesses, notes, sources, responsible agents, and portions of a base text.","TEI supplies multiple apparatus-linking methods and provenance-related attributes, making a machine-readable dependency ledger technically compatible with established edition data.","Identifying and grouping variations is explicitly not purely mechanical, so claim and edge coding necessarily contains scholarly judgment.","TEI does not prescribe a text-critical methodology or spectral review practice."]},{"source_id":"S4","title":"PROV-O: The PROV Ontology","publisher":"World Wide Web Consortium","url":"https://www.w3.org/TR/prov-o/","source_class":"STANDARD","publication_date":"2013-04-30","accessed_at":"2026-08-03","claims_supported":["PROV-O is a W3C Recommendation for interoperable provenance representation that can be specialized for application-specific domains.","It represents derivation, use, influence, revision, quotation, primary-source relations, responsibility, and provenance chains.","It provides a standards-based starting vocabulary for a dependency ledger but does not provide verification-queue ranking or spectral triage."]},{"source_id":"S5","title":"The Critical Apparatus Ontology (CAO): Modelling the TEI Critical Apparatus as a Knowledge Graph","publisher":"University and City Library of Cologne / Institut für Dokumentologie und Editorik","url":"https://kups.ub.uni-koeln.de/55230/","source_class":"PRIMARY_RESEARCH","publication_date":"2021","accessed_at":"2026-08-03","claims_supported":["A peer-reviewed study already proposes an OWL knowledge-graph representation of the TEI critical apparatus.","The work identifies expressive representation of scholarly views in critical apparatuses as a challenge.","CAO is close domain prior art for graph construction, although the source reports neither correction-propagation modes nor controlled verification-order comparisons."]},{"source_id":"S6","title":"Some Principles for Making Collaborative Scholarly Editions in Digital Form","publisher":"Digital Humanities Quarterly","url":"https://dhq.digitalhumanities.org/vol/11/2/000293/000293.html","source_class":"PRIMARY_RESEARCH","publication_date":"2017","accessed_at":"2026-08-03","claims_supported":["Textual Communities is an implemented collaborative scholarly-editing system with project-leader allocation and review of transcription work.","It supports document and intellectual-entity views, TEI compatibility, overlapping hierarchies, and embedded CollateX comparison.","The reported system had six major communities and about 150 active transcribers and editors, demonstrating credible organizational actors and digital workflow feasibility.","It is prior art for collaborative review and collation, not for claim-dependency propagation or spectral intervention selection."]},{"source_id":"S7","title":"NetworkX 3.6.1 Documentation: eigenvector_centrality","publisher":"NetworkX Developers","url":"https://networkx.org/documentation/stable/reference/algorithms/generated/networkx.algorithms.centrality.eigenvector_centrality.html","source_class":"OFFICIAL_PRODUCT_DOCUMENTATION","publication_date":"undated; version 3.6.1 current when accessed","accessed_at":"2026-08-03","claims_supported":["Commodity software computes weighted directed-network eigenvector centrality through power iteration.","For a strongly connected graph the relevant positive eigenvector is unique, while a non-strongly-connected graph may have multiple dominant eigenvectors and zero components.","The documentation exposes convergence failures and predecessor-versus-successor orientation choices that must be handled in the proposed ledger."]},{"source_id":"S8","title":"SciPy Manual: Sparse Eigenvalue Problems with ARPACK","publisher":"SciPy Community","url":"https://scipy.github.io/devdocs/tutorial/arpack.html","source_class":"OFFICIAL_PRODUCT_DOCUMENTATION","publication_date":"undated; SciPy 2.0.0.dev documentation when accessed","accessed_at":"2026-08-03","claims_supported":["Commodity numerical software can compute eigenpairs for real or complex nonsymmetric square matrices and can operate on sparse matrices or linear operators.","ARPACK can retrieve selected extremal eigenpairs without dense full decomposition.","The documentation warns that some requested eigenvalues can converge slowly or yield anomalous results, supporting explicit numerical diagnostics rather than unconditional trust in a modal basis."]}],"problem_evidence":{"support":"MODERATE","rationale":"Official and authoritative editorial sources visibly establish a consequential accuracy problem: editions require repeated independent checking, verification of annotations and quotations, sampled accuracy review, correction records, consistent apparatus, and referential integrity. TEI confirms that apparatuses contain structured links among text, readings, witnesses, notes, sources, and editorial responsibility. This supports the existence and importance of cross-element consistency work under limited staff time. However, none of the eight sources measures how often one corrected claim produces multi-hop downstream obligations, recirculating deficits, or stable spectral modes. The proposal's specific network-dynamic characterization therefore remains unverified.","source_ids":["S1","S2","S3"]},"stakeholder_evidence":{"support":"MODERATE","rationale":"Credible adopters and authorizers are identifiable: general/project editors and scholarly-edition vetters control review and acceptance, while independent editors verify annotations. MLA vetting questions and the documentary-editing guide express concrete demand for accuracy, correction control, provenance, and independent checks; Textual Communities demonstrates project-leader allocation and review in live collaborative editions. No source records an editor, press, MLA committee, or funder asking for spectral dependency triage specifically, and no pilot partner or data authorization has been secured.","source_ids":["S1","S2","S6"]},"prior_art":{"proximity":"ADJACENT_PRIOR_ART","closest_analogues":[{"name":"TEI critical-apparatus encoding plus conventional independent editorial verification","similarity":"Provides structured apparatus entries, text links, witness/source attribution, responsibility metadata, correction records, and established human checking workflows—the raw material and baseline workflow the proposal would use.","remaining_difference":"It does not encode a claim-level rechecking-propagation operator, calculate distributed modes, or prioritize a fixed review budget using held-out cascade outcomes.","source_ids":["S1","S2","S3"]},{"name":"Critical Apparatus Ontology (CAO)","similarity":"Represents the critical apparatus of a digital scholarly edition as an OWL knowledge graph and is the closest domain-specific graph analogue found.","remaining_difference":"The available study describes representation, not dependency weights, repeated deficit propagation, eigendecomposition, intervention sensitivity, residual queues, or comparative verification outcomes.","source_ids":["S5"]},{"name":"Textual Communities with CollateX","similarity":"An implemented system manages collaborative scholarly editing, assigns and reviews work, represents document/entity structure, and compares textual versions.","remaining_difference":"Its comparison and workflow features do not perform claim-dependency cascade prediction or modal review-order optimization.","source_ids":["S6"]},{"name":"W3C PROV-O provenance chains","similarity":"Supplies standardized derivation, influence, revision, quotation, source, agent, and responsibility relations suitable for a traceable dependency ledger.","remaining_difference":"It is domain-general provenance infrastructure and supplies neither an editorial dependency taxonomy nor any prioritization algorithm.","source_ids":["S4"]},{"name":"Eigenvector centrality and sparse eigensolvers","similarity":"Established software already computes dominant eigenvectors and eigenpairs on weighted directed or nonsymmetric graphs.","remaining_difference":"These tools establish computational availability, not the validity of interpreting editorial reliance as a stationary propagation process or the superiority of a modal verification queue.","source_ids":["S7","S8"]}],"distinctive_claim_remaining":"Within a preregistered, bounded critical-edition section that exhibits reproducible multi-hop claim dependencies, a stable and numerically well-conditioned distributed mode will predict blinded, held-out genuine rechecking obligations better than direct backlinks, reachability, degree, static PageRank-style prominence, and a flat queue; moreover, one-at-a-time mode-sensitive intervention ranking will find more adjudicated downstream inconsistencies within the same verification budget than the best of those comparators. The claim fails if the graph is shallow or effectively acyclic, coding is not reproducible, the modes are unstable or ill-conditioned, or the modal queue does not beat the best simple comparator.","confidence":"MODERATE"},"implementation_evidence":{"support":"MODERATE","rationale":"TEI, PROV-O, CAO, Textual Communities, NetworkX, and SciPy jointly show that structured apparatus data, provenance graphs, collaborative review workflows, eigenvector calculations, and sparse eigensolvers are technically available. A graph of at most 100 claims is computationally trivial. The hard feasibility questions are semantic and empirical: independently checkable claim boundaries, repeatable direction and weight coding, access to version history and sources, sufficient historical corrections, and blinded adjudication. There are also two mathematical hazards. First, a nonnegative column-normalized propagation matrix is norm-nonexpansive and generally cannot represent a genuinely growing deficit mode without an additional amplification parameter; persistence may partly be imposed by normalization. Second, a shallow directed acyclic graph has only zero adjacency eigenvalues, while a reducible or nonnormal graph may lack a unique, stable, interpretable eigenbasis. Strongly connected components, cycle depth, eigenvector condition numbers, pseudospectral sensitivity, and alternative formulations must therefore be checked before interpreting modes. Human authority, read-only operation, residual preservation, and re-estimation make a safe pilot plausible, but actual permission to use edition files, version history, licensed sources, and confidential disagreements is not externally established.","source_ids":["S3","S4","S5","S6","S7","S8"]},"scores":{"meaningful_impact":{"score":3,"rationale":"Reliable apparatus and annotation matter to scholarly editions, and official workflows invest substantial checking effort. The incidence and consequence of missed multi-hop conflicts are not quantified.","source_ids":["S1","S2","S3"]},"stakeholder_pull":{"score":3,"rationale":"Editors and vetters express strong demand for accuracy, independent verification, correction tracking, and referential integrity, but no adopter has requested this spectral mechanism.","source_ids":["S1","S2","S6"]},"incremental_advantage":{"score":2,"rationale":"The proposed comparison against backlinks, reachability, degree, static prominence, and flat review is meaningful, but no external evidence yet shows that modal ordering improves any editorial outcome.","source_ids":["S1","S5","S6","S7"]},"distinctiveness_plausibility":{"score":3,"rationale":"Knowledge-graph apparatuses and spectral algorithms already exist separately; outcome-tested modal intervention ordering appears contrastive in the searched evidence. Plausibility depends on finding reproducible cycles or persistent multi-hop structure rather than an effectively acyclic graph.","source_ids":["S5","S6","S7","S8"]},"technical_implementability":{"score":3,"rationale":"A 100-node ledger and its comparator algorithms are easy to compute with mature tools, but the proposed column normalization cannot generate amplification as stated, and directed nonnormal or reducible graphs can make full modal interpretation unstable.","source_ids":["S7","S8"]},"adoption_authority_feasibility":{"score":4,"rationale":"General or project editors already allocate review work, and independent checking is established practice. A read-only recommendation queue can remain subordinate to scholarly adjudication.","source_ids":["S1","S2","S6"]},"evidence_readiness":{"score":2,"rationale":"Standards and tools are ready, but the decisive evidence requires an authorized edition section, proprietary version history or editor-approved synthetic corrections, dual coding, and blinded adjudication.","source_ids":["S1","S2","S3","S6"]},"safety_net_benefit":{"score":4,"rationale":"A read-only pilot, human acceptance authority, comparator queues, a preserved residual queue, and automatic suspension on instability substantially limit harm. Confidentiality and automation-bias risks still require project-specific controls.","source_ids":["S1","S2","S3"]},"scalability":{"score":2,"rationale":"Computation scales readily, but claim individuation, dependency coding, source checking, and scholarly adjudication are labor-intensive and domain-specific. Existing collaborative systems show workflow scalability but not scalable claim-dependency mapping.","source_ids":["S6","S7","S8"]}},"score_confidence":"MODERATE","costs":{"first_evidence":{"band_2026_usd":"10K_TO_50K","scope":"Resource-equivalent cost for one read-only pilot of at most 100 claims: protocol design, two independent dependency maps, reconciliation, graph diagnostics, coding of historical or synthetic corrections, comparator evaluation, blinded scholarly adjudication, and a short report.","confidence":"LOW","assumptions":["Approximately 120-300 total hours across two mappers, a textual scholar/adjudicator, and a quantitative analyst.","Existing edition files, authorized sources, and version history are available without acquisition or licensing fees.","Open-source graph and numerical software is used.","This is a resource-equivalent estimate, not a vendor quote; no direct 2026 labor-rate benchmark was established by the eight sources."],"source_ids":["S2","S6","S7","S8"]},"initial_deployment_startup":{"band_2026_usd":"50K_TO_250K","scope":"Build a reusable ledger schema, TEI/PROV import and export, audit trail, access controls, comparator suite, diagnostics for reducibility and mode conditioning, review interface, documentation, and staff training for one edition project.","confidence":"LOW","assumptions":["One existing digital-edition platform and one bounded corpus are integrated.","No automated judgment or text rewriting is implemented.","A small software/data team works for roughly three to nine person-months.","Existing TEI or structured data substantially reduces transcription work."],"source_ids":["S3","S4","S5","S6","S7","S8"]},"operational_launch":{"band_2026_usd":"250K_TO_1M","scope":"Map and quality-control a substantial multi-section or volume-scale edition, integrate the tool into pre-freeze governance, conduct security and rights review, run repeated held-out validations, and support scholarly adjudication through launch.","confidence":"LOW","assumptions":["Several hundred to several thousand independently checkable claims require human review.","Two-person coding is retained for a validation sample rather than every obvious edge.","Licensed or confidential materials remain inside the project's existing controlled environment.","The range is dominated by scholarly labor and integration rather than eigendecomposition compute."],"source_ids":["S1","S2","S3","S6"]},"annual_recurring":{"band_2026_usd":"50K_TO_250K","scope":"Ongoing claim and edge maintenance, re-estimation after revisions, periodic validation, residual-queue review, access-control administration, training, and support for one active edition or series.","confidence":"LOW","assumptions":["One part-time technical maintainer and recurring textual-scholar/research-assistant effort are needed.","Update frequency follows substantive note or apparatus changes rather than continuous monitoring.","Open-source software avoids material license fees.","Manual scholarly adjudication remains mandatory."],"source_ids":["S1","S2","S6"]}},"verified_pipeline_gates":{"externally_supported_problem":{"status":"YES","reason":"Official and authoritative sources require repeated verification, independent checking, accuracy sampling, correction tracking, apparatus consistency, and referential integrity. They verify the broader problem, though not the prevalence of spectral or recirculating dependencies.","source_ids":["S1","S2","S3"]},"externally_credible_adopter_or_authorizer":{"status":"YES","reason":"Scholarly-edition vetters, general/project editors, and collaborative edition leaders are identifiable actors with authority to approve, allocate, review, and accept verification work.","source_ids":["S1","S2","S6"]},"distinct_testable_incremental_claim":{"status":"YES","reason":"The remaining claim specifies held-out cases, equal review budgets, blinded adjudication, named comparator queues, measurable ranking outcomes, and explicit failure conditions.","source_ids":["S1","S5","S6","S7"]},"bounded_next_evidence_step":{"status":"YES","reason":"A read-only, at-most-100-claim pilot with dual coding, historical or approved synthetic corrections, held-out evaluation, and fixed comparators is bounded in scope and reversible.","source_ids":["S1","S2","S6"]},"no_unresolved_safety_or_authority_stop":{"status":"UNCERTAIN","reason":"The proposed governance is appropriately human-controlled and read-only, but no actual editor authorization, repository permission, confidentiality plan, or lawful access to version history and licensed sources was externally verified.","source_ids":["S1","S2","S3"]},"credible_cost_scope_and_range":{"status":"UNCERTAIN","reason":"The scopes and labor drivers can be bounded, but the eight-source evidence set provides no direct 2026 wage, integration-cost, or edition-size benchmark. The bands are transparent resource-equivalent estimates with low confidence.","source_ids":["S2","S6"]}},"next_evidence_step":"With written authorization from a general or project editor, freeze one completed chapter containing no more than 100 independently checkable claims. Preregister the claim-boundary rules, edge taxonomy, weight schemes, review budgets k=10 and k=20, primary outcome (blinded-adjudicated genuine downstream inconsistencies found within k checks), secondary outcomes (precision@k, recall@k, mean rank of first genuine downstream inconsistency, and mapper hours), and a required superiority margin over the best comparator. Two mappers independently code the ledger; stop if agreement is below a preregistered threshold such as Cohen's kappa 0.70. Before spectral analysis, report strongly connected components, cycle depth, graph reducibility, spectral radius, eigenvalue multiplicity, eigenvector condition numbers, and sensitivity to plausible weight perturbations; stop if the graph is effectively acyclic, the operator is defective or ill-conditioned, or normalization mechanically creates the claimed persistence. Use temporally earlier adjudicated corrections for fitting and later corrections for held-out testing, or editor-approved synthetic source changes generated without reference to queue outputs. Compare flat, direct-backlink, transitive-reachability, degree, static PageRank-style, and modal-sensitivity queues under identical budgets. Adjudicators remain blinded to queue identity. Falsify the intervention claim if the modal queue fails to exceed the best comparator by the preregistered margin, reverses materially across plausible weight schemes, or leaves consequential structured residuals. Make no edition changes during the study.","blocking_evidence":["No published estimate was found for the prevalence, depth, or consequence of correction cascades among claims in critical-edition notes.","No edition partner, written authorization, repository permission, confidentiality plan, or lawful access to proprietary version history was verified.","No independently coded claim-dependency graph was available to establish inter-rater reliability, cycles, strongly connected components, or stable distributed modes.","No held-out adjudicated correction cases compare modal ordering with flat, backlink, reachability, degree, or static-prominence queues.","The stated nonnegative column-normalized operator cannot produce a genuinely growing total-deficit mode without an additional amplification or reproduction parameter; the exact dynamics require correction and preregistration.","A directed graph may be acyclic, reducible, nonnormal, or defective, making eigenvalue-based persistence trivial, nonunique, or numerically unstable; conditioning and perturbation evidence is absent.","No direct evidence supports the proposed 2026 labor and integration cost bands.","World novelty, patentability, freedom to operate, market size, and realized impact were not measured."],"research_disposition":"PARTNERED_RESEARCH_PROGRAM","world_novelty_boundary":"The search establishes only that apparatus verification, TEI encoding, provenance chains, apparatus knowledge graphs, collaborative collation workflows, eigenvector centrality, and eigensolvers are established or adjacent practices. It does not establish whether the specific combination of claim-level rechecking propagation, modal stability diagnostics, intervention sensitivity, residual preservation, and held-out editorial queue comparison has or has not appeared elsewhere. World novelty, patentability, freedom to operate, market size, and realized impact remain explicitly unmeasured.","arm":"COMPLETE_PROPOSAL_PORTFOLIO","candidate_version":0,"controller_recommendation":{"action":"STOP_EMPIRICAL_RESEARCH_NEEDED","repairable":false,"material_progress_observed":true,"progress_targets":["Secure written authorization and controlled access to one frozen chapter, its relevant sources, and adjudicated version history.","Demonstrate reproducible claim and edge coding at a preregistered agreement threshold.","Resolve the propagation definition so normalization does not mechanically preclude amplification or manufacture persistence.","Show nontrivial multi-hop or cyclic structure plus a stable, well-conditioned modal representation under resampling and edge-weight perturbation.","Beat flat, backlink, reachability, degree, and static-prominence comparators on blinded held-out cases under identical review budgets by a preregistered margin.","Verify that consequential low-loading cases remain protected by the residual queue and ordinary proofing.","Record actual mapper, adjudicator, analyst, integration, and maintenance hours to replace low-confidence cost estimates.","Complete project-specific confidentiality, repository-permission, licensing, and access-control review."],"reason":"Bounded web research supports the broader verification problem, credible editorial authority, adjacent graph and workflow prior art, and basic computational feasibility. It cannot determine whether real critical-edition claim graphs contain reproducible stable modes or whether modal sensitivity improves verification order. Those decisive facts require authorized proprietary data, dual human coding, scholarly adjudication, and offline comparative testing. Under the required controller rule, this empirical-research stop is not repairable within the present evaluation."},"proposal_index":2}