spb/wp44_uqo

1% ============================================================================2% Part III — Data and Systematic Review Protocol3% Written by Agent L (Part III instance), 2026-11-03; revised by Agent L/M (fixer) 2026-11-03 after Agent K critique (K-20 to K-36, corpus v3, A-006). Every corpus number is a macro from tables/numbers.tex4% (generated by tables/make_numbers.py from data/metadata/prisma_counts.json and linkage_report.json).5% Sources of truth: RESEARCH_PLAN.md §2–§3, protocol/PROTOCOL.md, protocol/amendments.csv,6% reports/PHASE0_FEASIBILITY.md, data/raw/README.md, ingestion/*.py, systematic_review/define_corpus.py.7% ============================================================================89\section{Information sources and conditions of access}10\label{sec:data-sources}1112A field-level mapping study stands or falls with the bibliographic databases on which it is built. The choice13of sources determines which works can be observed at all, how reliably they can be dated and typed, whether14citation links are available, and under what licence derived data may be redistributed. This part documents15these choices for WP44 and reports, stage by stage, how the mapping corpus was assembled from the raw16sources. The exposition follows the order of the pipeline (identification, linkage, deduplication,17eligibility, scope, dating) and closes with the quality indicators and the research-integrity procedures that18govern every later analysis. The reporting structure is modelled on the PRISMA 2020 items that apply to a19mapping study \citep{page2021prisma,page2021prismaexplanation}; items concerned with risk of bias and effect measures20have no counterpart in a bibliometric design and are replaced, as PRISMA-ScR does for scoping reviews21\citep{tricco2018prismascr}, by indicators of metadata quality and coverage (Section~\ref{sec:data-quality}).22The search-reporting recommendations of PRISMA-S \citep{rethlefsen2021prismas} are followed to the extent23that a database-scope strategy without keyword queries permits.2425\subsection{Why a multi-source design}2627No single database indexes computer science completely, and the gaps are not random. The venue-centred28coverage of DBLP, the publisher-centred coverage of Crossref, the preprint-centred coverage of arXiv and the29aggregating coverage of OpenAlex each miss a distinct part of the literature and type the part they do cover30in distinct ways. Comparative studies of bibliographic data sources have documented large differences in31coverage between Web of Science, Scopus, Dimensions, Crossref and Microsoft Academic32\citep{visser2021comparison,martinmartin2021coverage}, and the journal-level coverage of the two commercial33indexes is known to under-represent conference proceedings, which carry a large share of computer-science34output \citep{mongeon2016coverage}. For the open successor of Microsoft Academic, OpenAlex, the first35independent assessments report strong coverage of works with DOIs together with weaker and uneven metadata on36affiliations, document types and references \citep{alperin2024openalex,culbert2025openalexcoverage,37scheidsteger2023magopenalex,delgadoquiros2024completeness}. These findings motivate the design adopted here: one38source supplies the backbone (OpenAlex), a second supplies the definition of what counts as computer science39(DBLP), a third supplies the preprint layer and its dates (arXiv), and a fourth checks the coverage of a40subfield with a dense, curated venue structure (the ACL Anthology). Crossref is used for verification only.4142\subsection{OpenAlex}4344OpenAlex \citep{priem2022openalex} is the backbone of the corpus. It supplies, for every work, the45identifiers (OpenAlex~W identifier, DOI, arXiv identifier where present), the publication date and type, the46title and an inverted-index abstract where available, the venue (source) and its type, the authorships with47institutions and countries, the topic assignment at three levels (field, subfield, topic) with a primary48topic, the list of referenced works (the citation graph), the open-access status and licence, and the49field-weighted citation impact together with yearly citation counts. OpenAlex is released under the CC050public-domain dedication, which permits redistribution of record-level derived data.5152Two access routes exist and both were used, for different purposes. The public snapshot is a versioned bulk53release in gzipped JSON Lines, published in a public S3 bucket; the release used here is dated542026-10-14 and comprises 1{,}898 part files for the \emph{works} entity, 675.3~GB compressed, with a manifest55listing the record count of each file. The snapshot was streamed from the primary compute node, each file's56content length checked against the manifest and a SHA-256 of the compressed bytes recorded in a SQLite ledger57before parsing; every one of the 1{,}898 files parsed without error and the sum of the per-file record counts58equals the manifest total of \LinkSlimTotal{} works. Two outputs were written during this single pass. The59first is the full JSON record of every work carrying \emph{any} topic in OpenAlex field~17 (Computer Science)60with a publication year between 2018 and 2026, stored as Zstandard-compressed JSON Lines and partitioned by61the snapshot's own date partitions (\PrismaIdentifiedOpenalexPassOneCsTopicFilter{} records in this first62pass, about 20~GB; a second pass by identifier list, described in Section~\ref{sec:data-linkage}, later added63\PrismaIdentifiedOpenalexPassTwoDblpArxivLinked{} records). The second is a \emph{slim}64table of every work in the snapshot (identifier, DOI, year, type, primary topic, subfield, field and domain,65citation count and a few flags), kept in Parquet so that works cited by computer-science papers but published66in other fields can be resolved for the interdisciplinary analyses of Part~VIII (decision D2 of67\texttt{RESEARCH\_PLAN.md}). Records flagged by OpenAlex as XPAC expansion records are retained at this stage68with their flag so that the exclusion decision can be taken, and counted, later (Section~\ref{sec:data-eligibility}).69The smaller entity snapshots (topics, subfields, fields, domains, sources, institutions, publishers, funders,70concepts) were downloaded the same day with a per-file SHA-256 ledger; author entities were not downloaded71because the authorship data embedded in each work record suffice for the analyses planned.7273The second route is the REST API. Without an API key the public endpoint operates under a shared daily budget74of USD~0.10, list-and-filter calls are priced at USD~0.0001 each, and cursor paging is unavailable, so that a75single filtered query cannot return more than 10{,}000 results through basic paging. Single-record lookups by76identifier are free. The API was therefore used for two narrowly defined tasks. First, an incremental harvest77of works with any computer-science topic and a publication date between 2026-10-01 and 2026-11-02, intended78to capture records added after the snapshot was cut; the harvest is sliced by publication day and, where a day79exceeds the paging limit, split into records with a computer-science primary field and records without one.80Raw JSON pages are stored verbatim under a dated directory with a ledger of counts and cost, and the job81resumes on the following day when the budget is exhausted. Second, free single-record lookups by DOI serve as82one of the verification routes for the bibliography of this manuscript (Section~\ref{sec:data-integrity}).83A free API key, which would raise the daily budget, requires the principal investigator to act on an e-mail84and was deliberately not obtained without instruction; the mandate forbids assuming credentials that have not85been tested.8687\subsection{DBLP}8889The DBLP computer science bibliography \citep{ley2002dblp,ley2009dblp} is a curated, venue-centred index that90has for three decades defined, in practice, the set of journals and conference series regarded as computer91science. It is used here in that capacity: a work indexed by DBLP is in computer science by construction, and92DBLP venue streams feed the taxonomy of Part~IV. The complete XML dump (\texttt{dblp.xml.gz}, dated 2026-11-029322:39~UTC, 1.12~GB) was downloaded together with its DTD and MD5 checksum, the checksum verified, and the file94parsed with a validating XML parser into one row per publication record (articles, conference papers,95proceedings volumes, books, book chapters, doctoral and master's theses, data records; the \texttt{www}96person records are skipped). The whole archive was kept so that per-year counts are available for the97historical baseline and for context. The DBLP dump is released under the Open Data Commons Attribution licence98(ODC-BY~1.0). DBLP contributes, in addition to the scope definition, author person identifiers, venue streams,99and DOI links for most records; the \texttt{ee} fields were also scanned for arXiv URLs, with a result reported100in Section~\ref{sec:data-linkage}.101102\subsection{arXiv}103104The arXiv e-print repository \citep{ginsparg2011arxiv} is the main preprint channel of computer science and,105for the period under study, the place where a large share of the most cited work first appeared. Its106metadata are harvested through the OAI-PMH interface (\texttt{set=cs}, \texttt{metadataPrefix=arXiv}) in107monthly datestamp windows from January 2018 to 2 November 2026. The harvester waits at least 3.5~s between108requests and honours \texttt{Retry-After} headers; every raw XML page is kept verbatim as the immutable109snapshot and parsed afterwards into a single Parquet table. Two properties of OAI-PMH matter for what follows.110The \texttt{from}/\texttt{until} parameters filter on the record's \emph{datestamp}, the time of its last111update, so that a monthly window contains every record updated in that month, not every record first112submitted then; duplicates across windows are resolved at parse time with the latest datestamp winning. The113\texttt{created} element gives the first submission date, which is the date used throughout this study for114preprints. The harvest returned \LinkArxivCsRecords{} distinct records, of which115\LinkArxivCreatedTwoZeroOneEightTwoZeroTwoSix{} were first submitted in 2018--2026; the remainder are older116e-prints updated during the window. The arXiv metadata are released for reuse under the repository's metadata117terms; e-prints themselves are governed by per-paper licences and are downloaded, for the high-priority set118only, within those licences and at the repository's rate limits. The use of arXiv as a research dataset, and119the caveats attached to its category labels and versioning, are discussed by \citet{clement2019arxiv}.120121\subsection{ACL Anthology}122123The ACL Anthology is the curated archive of the computational-linguistics community and has itself served as a124bibliometric corpus \citep{radev2013aclanthology}. Its BibTeX export with abstracts (\texttt{anthology+abstracts.bib.gz},12546.7~MB, SHA-256 recorded) was downloaded on 2026-11-02 and parsed into a table of \AclParsedEntries{} entries, of which126\LinkAclRecords{} are conference papers or journal articles and \LinkAclTwoZeroOneEightTwoZeroTwoSix{} fall in1272018--2026. The Anthology plays a diagnostic role: because its venue structure is complete and hand-maintained,128the share of its entries that the backbone can be matched to is a direct measure of OpenAlex coverage for a129subfield that publishes overwhelmingly in conference proceedings. It also supplies abstracts for those entries130in which OpenAlex has none. The export is licensed CC~BY~4.0.131132\subsection{Crossref, Semantic Scholar and other sources}133134Crossref \citep{hendricks2020crossref} is the registration agency behind most DOIs in the corpus and offers135open metadata through a REST API with a polite pool. It is not used to build the corpus, whose backbone136already carries Crossref-derived metadata through OpenAlex, but for three verification purposes: resolution of137every DOI cited in this manuscript, a cross-check of publisher-assigned document types on samples, and a138cross-check of citation counts (\texttt{is-referenced-by-count}) on a sample of highly cited works in139Part~IX.140141The Semantic Scholar Academic Graph \citep{kinney2023s2ag} was listed in the mandate among the priority sources142because of two signals it alone provides, the influential-citation classification and citation intent. At the143time of the feasibility audit, anonymous requests to its API returned HTTP~429 with an instruction to apply for144a key, and the bulk-dataset endpoint requires a key outright. In keeping with the rule that inaccessible data145be recorded rather than circumvented, Semantic Scholar is recorded as unavailable (decision D3) and no analysis146in this manuscript uses influential-citation or citation-intent signals; where Part~VIII discusses the147strength of citation links, it does so with OpenAlex reference lists and the knowledge-graph extraction of148Part~IV, whose limits are stated there. The Microsoft Academic Graph \citep{wang2020mag}, from which OpenAlex149inherited much of its historical coverage, was discontinued at the end of 2021 and is not an independent source150for the period studied. Kaggle credentials for the arXiv snapshot distributed there were absent and were not151needed, OAI-PMH being sufficient. Benchmark repositories and code-hosting platforms enter only in the152reproducibility and benchmark analyses scoped in Part~IV and are not part of the bibliographic corpus.153154Table~\ref{tab:data-sources} summarises the sources, the access route, the licence and the role of each.155156\begin{table}[htbp]157\centering\footnotesize158\caption{Information sources of the WP44 corpus (all accessed 2026-11-02/03).}159\label{tab:data-sources}160\begin{tabularx}{\textwidth}{>{\raggedright\arraybackslash}p{2.5cm}>{\raggedright\arraybackslash}X>{\raggedright\arraybackslash}p{2.1cm}>{\raggedright\arraybackslash}X}161\toprule162Source & Access and snapshot & Licence & Role \\163\midrule164OpenAlex works snapshot & public S3 bucket, release 2026-10-14, 1{,}898 JSONL files, \LinkSlimTotal{} works & CC0 & backbone: identifiers, dates, types, topics, authorships, references, abstracts, OA status \\165OpenAlex REST API & no key; USD~0.10 per day shared budget; increment 2026-10-01 to 2026-11-02; single lookups & CC0 & post-snapshot increment; reference verification \\166DBLP & XML dump 2026-11-02 22:39~UTC, MD5 verified & ODC-BY~1.0 & computer-science scope; venue streams; author identifiers \\167arXiv & OAI-PMH set \texttt{cs}, 2018-01 to 2026-11, raw pages kept & arXiv metadata terms & preprint layer; first-submission dates; categories \\168ACL Anthology & BibTeX export with abstracts, 2026-11-02 & CC~BY~4.0 & coverage check for NLP venues; abstracts \\169Crossref & REST API, polite pool & open metadata & verification of DOIs, types and citation counts \\170Semantic Scholar & API key required; anonymous calls refused & --- & \emph{unavailable, recorded} \\171\bottomrule172\end{tabularx}173\end{table}174175\section{Identification: what each source contributes}176\label{sec:data-identification}177178There is no keyword search in this study. A mapping of an entire discipline cannot be defined by a query179string without importing the vocabulary of whoever wrote it; it is defined instead by database scope filters180that are themselves documented and reproducible, which is the strategy recommended for mapping studies in181software engineering and adopted here for computer science as a whole182\citep{petersen2008mapping,petersen2015mapping}. The identification stage (S0 in the notation of the research183plan) therefore consists of four scope filters, one per source, applied to immutable snapshots whose dates and184hashes are recorded in \texttt{data/raw/README.md}.185186\subsection{Scope filters}187188The OpenAlex filter retains every work with at least one topic, primary or secondary, in field~17 and a189publication year in $[2018, 2026]$. The choice of \emph{any} topic rather than the primary topic is deliberate190and conservative at this stage: a work whose primary topic lies in, say, statistics or electrical engineering191but which carries a secondary computer-science topic is a candidate, and the decision whether it belongs to192the core or to the penumbra is deferred to the eligibility stage, where it is made explicit and counted193(Section~\ref{sec:data-core}). The filter yields194\PrismaIdentifiedOpenalexPassOneCsTopicFilter{} records in the first pass over the snapshot; a second pass by195identifier list (Section~\ref{sec:data-linkage}) adds \PrismaIdentifiedOpenalexPassTwoDblpArxivLinked{}196records that carry no computer-science topic but are reached through DBLP and arXiv identifiers, so that197\PrismaIdentifiedOpenalexTotalRecordsLoaded{} OpenAlex records are loaded in all. The DBLP filter retains the publication records of type article, inproceedings, incollection, book,198phdthesis and data with a year in $[2018, 2026]$: \PrismaIdentifiedDblpTwoZeroOneEightTwoZeroTwoSix{} records.199The arXiv filter retains the records of set \texttt{cs} whose first submission falls in the window:200\PrismaIdentifiedArxivCsTwoZeroOneEightTwoZeroTwoSix{} records. The ACL filter retains the conference papers201and journal articles of the Anthology with a year in the window:202\PrismaIdentifiedAclTwoZeroOneEightTwoZeroTwoSix{} records. The four counts overlap heavily, since most DBLP203and ACL records and a majority of arXiv records also exist in OpenAlex; the overlap is resolved by record204linkage, not by summation, and no total of identified records is reported as if the sources were disjoint.205206\subsection{The OpenAlex topic classifier as a scope instrument}207208Because the OpenAlex filter relies on an automated topic classifier, its properties bear on everything209downstream. OpenAlex assigns to each work up to three topics from a fixed list of 4{,}516, organised in 252210subfields, 26 fields and four domains; the assignment is made by a model applied to the title, abstract and211venue, and the field label of a work is inherited from its topics rather than assigned directly. Two212consequences follow. A work can be labelled computer science because one of its topics belongs to field~17213even when its content is only marginally computational, and conversely a work in a computer-science venue can214receive no computer-science topic when its abstract is missing or its vocabulary is domain-specific. The first215error inflates the candidate set with what will be called the penumbra; the second removes genuine216computer-science works from the first-pass candidate set. DBLP is used to correct the second error, which is217why the linkage stage has two passes (Section~\ref{sec:data-linkage}). The literature on OpenAlex has so far218examined its coverage and metadata completeness rather than the accuracy of its topic labels219\citep{alperin2024openalex,delgadoquiros2024completeness}; the present study treats the labels as a scope220instrument whose errors are bounded by the DBLP and arXiv memberships and quantified in221Section~\ref{sec:data-quality}, not as ground truth.222223\subsection{Verification of the scale against the mandate}224225The mandate set planning targets of more than one million candidate records and more than one hundred226thousand deeply classified records, subject to availability. The feasibility audit established, before any227ingestion, that the OpenAlex API alone reported more than six million works with a computer-science primary228field in 2018--2026 and that the DBLP dump held more than four million records in the same years; the targets229were therefore attainable without recourse to any restricted source. The audit also established what was230\emph{not} attainable: a Semantic Scholar key, an OpenAlex key, Kaggle credentials and the AWS command-line231tool were all absent, and the ingestion was designed around their absence (HTTPS streaming of the public232bucket, basic paging of the public API). The record of this audit is \texttt{reports/PHASE0\_FEASIBILITY.md}.233234\section{Record linkage across sources}235\label{sec:data-linkage}236237The four candidate sets describe overlapping populations with different identifiers. Record linkage (stage S1)238establishes, for each DBLP, arXiv and ACL record, the OpenAlex work it describes, so that one node per work239can be formed with the union of the metadata. The methods are deterministic and ordered by reliability; the240probabilistic framework of \citet{fellegi1969linkage} is not needed because the available keys (DOI, arXiv241identifier) are close to exact, and the residual title-based step is restricted to blocks where it is242unambiguous. The implementation follows the standard decomposition of data matching into key normalisation,243blocking, comparison and classification \citep{christen2012matching}; Part~IV gives the formal statement of the244keys and of the precision-estimation design, and this section reports what was done and what it produced.245246\subsection{Keys and matching order}247248All keys are normalised before comparison. DOIs are lower-cased. arXiv identifiers are lower-cased and249stripped of their version suffix (\texttt{v2}, \texttt{v3}, \ldots), so that all versions of an e-print map to250one identifier. Titles are reduced to a normalised string by lower-casing and deleting every character that is251not a letter or digit; the normalised title is used only when it has at least 25 characters, is unique within252its source, and is unique within OpenAlex, which is the blocking rule that keeps the title step from253producing spurious links between generic titles.254255For DBLP, the order is: (i) DOI equality, (ii) arXiv identifier equality, restricted to records not yet linked256on either side, (iii) normalised title equality with a year difference of at most one, restricted to records257not yet linked on either side and, when a first author is recorded in both sources, to pairs whose258first-author surnames agree. The DOI step linked \LinkLinkDblpOaDoi{} records and the title step259\LinkLinkDblpOaTitleYear{}; the arXiv step linked none, because the extraction of arXiv URLs from the DBLP260\texttt{ee} fields returned no candidate (\LinkDblpWithArxiv{} records with an extracted identifier). The261\texttt{ee} fields of this dump evidently do not carry \texttt{arxiv.org/abs} URLs in the form the extractor262expects; whether the CoRR entries are reached by the DOI step instead depends on DBLP's current conventions for263the \texttt{ee} field, which the audit did not characterise, and the question is left to the linkage-precision264review. The step is kept in the pipeline because other dumps may differ. The linkage is run on the union of265the two snapshot passes, so that every count in this section is a count after both passes. In total266\LinkDblpLinkedDistinct{} of the \LinkDblpTwoZeroOneEightTwoZeroTwoSix{} DBLP records in the window were267linked to an OpenAlex work, leaving \LinkDblpUnlinked{} unlinked. The composition of the unlinked set is268informative: \LinkDblpUnlinkedByTypeArticle{} journal articles, \LinkDblpUnlinkedByTypeInproceedings{}269conference papers, \LinkDblpUnlinkedByTypePhdthesis{} doctoral theses, \LinkDblpUnlinkedByTypeData{} data270records, \LinkDblpUnlinkedByTypeIncollection{} chapters and \LinkDblpUnlinkedByTypeBook{} books. A DBLP271record whose DOI exists in the slim table of all OpenAlex works but not among the loaded records describes a272work that is in OpenAlex but was not caught by the computer-science topic filter; this is the second error of273the topic classifier described above and the reason for the second pass. After that pass,274\LinkDblpUnlinkedWithDoiInSlim{} unlinked DBLP records are still in this situation; their full records were275not returned by the identifier scan, for a reason the pipeline has not characterised, and the list is kept276for the linkage-precision review.277278For arXiv, the order is: (i) arXiv identifier equality against the identifier OpenAlex records for the work,279(ii) DOI equality for e-prints that carry a journal DOI, (iii) normalised title equality with a year280difference of at most two, the wider tolerance reflecting the typical lag between a preprint and its published281version \citep{lin2020arxivcs}. The steps linked \LinkLinkArxivOaArxivId{}, \LinkLinkArxivOaDoi{} and282\LinkLinkArxivOaTitleYear{} records respectively, for \LinkArxivLinkedDistinct{} distinct e-prints linked.283For the ACL Anthology, the order is: (i) DOI equality, (ii) equality between the Anthology identifier and the284identifier embedded in an \texttt{aclanthology.org} landing-page URL among the work's OpenAlex locations,285(iii) normalised title equality with a year difference of at most one. The steps linked286\LinkLinkAclOaDoi{} records by DOI and \LinkLinkAclOaTitleYear{} by title; the URL step contributed no link287in this snapshot, and is kept for the same reason as the DBLP arXiv step. The result is \LinkAclLinkedDistinctTwoZeroOneEightTwoZeroTwoSix{} distinct 2018--2026 entries linked out of288\LinkAclTwoZeroOneEightTwoZeroTwoSix{}; the residual is examined in Section~\ref{sec:data-quality} as a289measure of OpenAlex coverage of conference-heavy subfields.290291\subsection{The second snapshot pass}292293A DBLP record whose DOI resolves to an OpenAlex work outside the topic filter is, by the scope rule, a294computer-science work that the first pass did not load. The same situation arises for arXiv e-prints whose295journal DOI exists in OpenAlex without a computer-science topic. The linkage script therefore writes the list296of OpenAlex identifiers that are needed but not loaded, from both routes. A second pass over the works297snapshot, restricted to this identifier list, retrieves the full records of these works298(\PrismaIdentifiedOpenalexPassTwoDblpArxivLinked{} records); they are normalised into the same Parquet299tables with a \texttt{\_p2} suffix, the DuckDB views are rebuilt over the union of both passes, and linkage300is re-run so that the DBLP-linked set gains the works that the topic classifier had missed. Rebuilt on the301union, the list of identifiers still needed shrinks to \LinkNeededOaIdsSecondPassDblpDoi{} from DBLP DOIs and302\LinkNeededOaIdsSecondPassArxivDoi{} from arXiv DOIs: every e-print whose journal DOI exists in OpenAlex was303reached, and the DBLP residual is the one discussed above. The second pass is what makes the DBLP membership304rule of Section~\ref{sec:data-core} operational: without it, ``DBLP-indexed'' would silently mean305``DBLP-indexed and labelled computer science by OpenAlex'', and the DBLP correction would be void precisely306where it matters. Because the second pass loads records by identifier without any restriction on their year,307it also admits a small number of works dated outside the window, which the eligibility stage removes under308code E1. DBLP records with no OpenAlex counterpart after both309passes (\PrismaLinkedDblpOnlyNoOpenalexRecord{} records) are kept in a separate table310with their year and type and are reported as such; they enter the volume counts of Part~IX only in the311sensitivity analysis that adds DBLP-only records, because they lack the abstracts, references and affiliations312on which every other indicator depends. Most of them are journal articles and conference papers with a DOI313that OpenAlex does not hold at all, together with doctoral theses, which DBLP indexes from university314repositories that OpenAlex covers unevenly.315316\subsection{Precision of the linkage}317318Exact-key links are correct whenever the key is correct in both sources; the known failure modes are DOIs319reassigned by publishers between versions, DOIs registered for a proceedings volume rather than a paper, and320e-prints whose OpenAlex record points to a different version. Title-based links can fail on near-identical321titles of distinct works (extended abstracts and full papers, workshop and conference versions). The protocol322pre-specifies the estimation of linkage precision on a sample of 500 DBLP--OpenAlex pairs stratified by method,323with automatic agreement checks (normalised-title similarity, year difference, first-author surname) and a324subsample of two pairs in five reserved for the principal investigator (checkpoint H-1). The sampling script325draws 250 DOI links, 200 title links and 50 arXiv links with a fixed seed; since the arXiv step produced no326links in this dump, the realised sample (\texttt{data/metadata/linkage\_sample.csv}) has 450 pairs, of which327180 are reserved for the principal investigator, and the precision estimate for the arXiv step is undefined328rather than reported as perfect. The automatic check passes 235 of the 250 DOI pairs (0.94) and 200 of the329200 title--year pairs (1.00). The fifteen DOI pairs that fail the automatic check are not fifteen wrong links:330inspection of the sample shows that most fail on a title that OpenAlex truncated at the subtitle, on a331first-author name written in a different order or without diacritics, or on a one- or two-year difference332between the DBLP year and the OpenAlex year; a few pairs, however, carry substantively different titles under333the same DOI, which is the signature of a retitled version or of a DOI that resolves to a different item than334the one DBLP describes, and these are genuine candidates for error. The DOI stratum is therefore not335error-free, contrary to what an exact-key method would suggest, and the human verdicts on the reserved pairs336are what will settle the stratum precision. At the time of writing no verdict has been entered and the337checkpoint is unsigned; the stratum-weighted estimate of Part~IV will be reported in Part~IX and Appendix~B338once it is, and every conclusion that depends on the title-based links is labelled as pending human review339until H-1 is signed.340341\section{Deduplication of preprint and published versions}342\label{sec:data-dedup}343344OpenAlex creates distinct work records for an arXiv e-print and for the journal or conference version of the345same paper whenever the two carry different DOIs, which is the normal case since arXiv registers DataCite DOIs346for its e-prints. In a corpus where preprinting is widespread this is the dominant form of duplication, and it347biases every count in the direction of fields that preprint most: a paper in machine learning that appears on348arXiv and then at a conference would be counted twice, a paper in a theory journal once. OpenAlex itself links349some preprint--publication pairs, but the links are incomplete, and prior work on discovering such links has350shown that neither DOIs nor titles alone suffice \citep{cabanac2021preprint}. Stage S2 therefore merges351preprint and published versions into one node per work, in the sense required by the mandate (one node per352\emph{work}, with both dates retained).353354\subsection{Pairing rule}355356The rule is conservative by design, favouring a small number of certain merges over a larger number of357probable ones. A \emph{preprint-like} record is one whose OpenAlex type is \texttt{preprint} or whose DOI is358an arXiv DataCite DOI (prefix \texttt{10.48550/arxiv}); a \emph{published-like} record is any other record359that is itself an admissible twin: dated within the window, not an XPAC expansion record, titled, and of a360research document type or DBLP-linked. The twin conditions are imposed so that a preprint is never merged361into a record that the eligibility stage would then remove, which would lose the work altogether. A362preprint-like and a published-like record are paired by either of two routes: (i) identical normalised titles363of at least 25 characters with publication years differing by at most two, or (ii) a shared arXiv identifier,364taken from the arXiv linkage where available and from the OpenAlex record otherwise. When a published record365matches several preprints, the earliest preprint is kept as its preprint; when a preprint matches several366published records, the earliest published record is kept; all other pairs are discarded, so that the pairing367is one-to-one. The pairing is confined to the loaded OpenAlex records; preprints that are not computer-science368candidates cannot be paired, which is immaterial since they would not enter the corpus. No fuzzy title369similarity is used at this stage, so that a preprint whose title changed at publication is merged only when370the second route identifies it; the consequence, a residual double count, is discussed below.371372\subsection{What is kept from each member of a pair}373374The published record is the surviving node. The preprint record receives the flag \texttt{duplicate\_of} with375the identifier of the published work and is excluded with reason code E5 (Section~\ref{sec:data-eligibility}).376The published node receives \texttt{preprint\_work\_id} and, through the arXiv linkage, the first-submission377date of the e-print, so that for every work the \emph{first date} is defined as the earliest of its OpenAlex378publication date, the publication date of its preprint record and the arXiv \texttt{created} date, with one379exception stated in Section~\ref{sec:data-dates}: a publication date of the form \emph{YYYY}-01-01 is treated380as a year-only date and is overridden by the preprint or e-print date when one exists. This first381date is the date variable of all monthly series in Part~IX; the publication date of the version of record is382kept alongside it. The distinction matters for structural-change analysis: dating a paper by its journal issue383would displace much of the research activity of 2023 into 2024 and 2025 and would blur any monthly series384around the November 2022 boundary, whereas the first date places the work at the moment it entered the public385record. OpenAlex attributes to the preprint record the citations that cite the preprint DOI; the impact386analyses (\texttt{bibliometrics/03\_citation\_impact.py}) therefore take the yearly citation counts of the387surviving node as the union of its own counts and those of its merged preprint record, so that a work whose388preprint remained the cited version is not under-counted. The merge is exact only for pairs the pairing found;389for the residual unmatched preprints discussed below, the citations stay divided between two nodes. Part~IV390states the resulting measurement assumption and Part~IX reports, for the high-priority set, the share of391citations that accrue to preprint records.392393\subsection{Outcome and residual duplication}394395The pairing identified \PrismaExcludedEFiveDuplicate{} preprint records as duplicates of a published work,396of which \PrismaExcludedSequentialEFiveDuplicate{} remained after the earlier exclusions had been applied.397Within the eligible corpus, \PrismaVersionRelationshipTitleYear{} works carry a preprint merged by the398title route and \PrismaVersionRelationshipArxivId{} by the shared-identifier route (a pair that satisfies399both is counted once, under either route). Set against the \PrismaEligibleByTypePreprint{} preprints that remain in the eligible corpus as works with no400published counterpart, the figure shows that most e-prints in the candidate set were not matched to a401published version within OpenAlex. Three explanations coexist and cannot be separated without manual review:402many e-prints are genuinely unpublished, consistent with the publication rates reported for computer-science403preprints \citep{lin2020arxivcs}; some published versions carry a modified title; and some published404versions are not in OpenAlex or lack a computer-science topic. The first explanation is benign; the second and405third leave a residual double count whose size is bounded in Part~IX by re-running the pairing with a title406similarity threshold rather than exact equality and reporting the difference as a sensitivity, not as a407correction. Because arXiv-linked works carry their e-print identifier, the sensitivity analyses can also408remove every preprint node whose e-print has a journal DOI in the arXiv metadata, whether or not the pairing409found the published record.410411\section{Eligibility criteria and exclusions}412\label{sec:data-eligibility}413414Stage S3 applies the eligibility criteria of the preregistered protocol (\texttt{protocol/PROTOCOL.md}, P2) to415every loaded record and records, for each, a set of Boolean exclusion flags with reason codes. The flags are416computed independently so that each code can be counted on its own and also sequentially, in the order417E1--E7, so that the PRISMA flow (Figure~\ref{fig:prisma}) sums correctly. Reporting both views is deliberate:418the independent counts show how large each problem is in the raw data, the sequential counts show how many419records each criterion removed given those before it.420421\subsection{Reason codes}422423\textbf{E1, outside window.} The publication date is missing, earlier than 2018-01-01 or later than4242026-11-02. Because the first-pass filter already restricted the publication year, this code catches, among425first-pass records, only those with a missing date or a date beyond the extraction date; it also removes the426second-pass records, loaded by identifier without a year restriction, whose publication date lies outside the427window. Forward-dated records, that is works whose428publication date lies in the future relative to the snapshot, are a documented feature of OpenAlex, where429publishers register articles for forthcoming issues; the window end removes those dated after the extraction430date, and the remaining 2026 records are discussed in Section~\ref{sec:data-dates}.431\PrismaExcludedEOneOutsideWindow{} records are flagged.432433\textbf{E2, non-research document type.} The protocol lists the document types that constitute research434output: article, conference paper, preprint, book chapter, book, dissertation, report, review and software435paper. The exclusion operates asymmetrically. A record that is \emph{not} DBLP-linked is excluded when its436OpenAlex type is outside the list. A record that \emph{is} DBLP-linked is kept whatever its OpenAlex type,437unless that type belongs to a hard-exclusion list of types that are never research outputs (editorial,438erratum, retraction notice, peer review, paratext, conference abstract, reference entry, dataset, software,439supplementary material, library guide, letter, grant, standard). The asymmetry reflects amendment A-001 and440is explained below. Independently of order, \PrismaExcludedETwoType{} records carry this flag. The largest441contributors are records typed as datasets (\PrismaExcludedETwoByTypeDataset{}), followed by the442residual type \texttt{other} (\PrismaExcludedETwoByTypeOther{}), library guides443(\PrismaExcludedETwoByTypeLibguides{}), software (\PrismaExcludedETwoByTypeSoftware{}), conference444abstracts (\PrismaExcludedETwoByTypeConferenceAbstract{}) and peer reviews445(\PrismaExcludedETwoByTypePeerReview{}); paratext, reference entries, editorials, supplementary materials,446errata, standards, book reviews, retraction notices, data papers and letters make up the remainder. The447prominence of datasets and software among records carrying a computer-science topic is itself a finding about448the backbone: OpenAlex indexes the DataCite registry, where repositories mint DOIs for research objects that449are not publications, and the topic classifier labels them by their descriptions.450451\textbf{E3, XPAC expansion records.} OpenAlex marks as XPAC the records it added from non-Crossref sources in452its coverage expansion; these carry sparser metadata and a higher share of non-research objects. The453protocol excludes them, and \PrismaExcludedEThreeXpac{} records carry the flag independently. Amendment A-001454introduced one exception: an XPAC record that is linked to a DBLP record is a computer-science publication455whatever its provenance in OpenAlex, and it is kept with the flag \texttt{xpac\_kept\_dblp}456(\PrismaExcludedXpacKeptDblp{} records). Sequentially, after the type criterion,457\PrismaExcludedSequentialEThreeXpac{} records are removed under E3, far fewer than the independent count458because most XPAC records are also non-research types.459460\textbf{E4, no title.} Records with a missing or empty title cannot be linked, deduplicated or embedded and are461excluded: \PrismaExcludedEFourNoTitle{} independently, \PrismaExcludedSequentialEFourNoTitle{} sequentially.462463\textbf{E5, duplicate version.} The preprint member of a merged pair (Section~\ref{sec:data-dedup}):464\PrismaExcludedEFiveDuplicate{} independently, \PrismaExcludedSequentialEFiveDuplicate{} sequentially.465466\textbf{E6, retracted.} Works flagged \texttt{is\_retracted} by OpenAlex, which draws on the Retraction Watch467data now distributed through Crossref, are \emph{not} removed from the mapping corpus: they are part of the468published record and of its volume, and their presence is itself an object of study given the record numbers469of retractions reported for 2023 \citep{vannoorden2023retractions}. They are kept with the flag and excluded470from every impact analysis (citation rankings, landmark selection, field-normalised scores).471\PrismaExcludedESixRetracted{} loaded records carry the flag, of which472\PrismaEligibleMembershipRetractedFlagged{} are in the eligible corpus.473474\textbf{E7, generic deposit platform.} Introduced by amendment A-002, this code excludes records whose primary475source is a generic deposit platform (Zenodo, Figshare, the Japanese Institutional Repositories DataBase,476LA~Referencia, Dialnet, the DOAJ-as-source entries and the OSTI OAI feed) unless the record is DBLP- or477arXiv-linked. Independently of order, \PrismaExcludedESevenGenericDeposit{} records carry the flag;478sequentially, after the five preceding criteria, \PrismaExcludedSequentialESevenGenericDeposit{} records are479removed under this code.480481One change of numbering must be recorded. In the frozen protocol, code E7 designated the penumbra of works482with a secondary computer-science topic only; the implementation reuses the number E7 for the generic-deposit483exclusion of amendment A-002 and carries the penumbra as a separate flag, \texttt{cs\_penumbra}. The484penumbra is not an exclusion in the same sense as E1--E7: it is a scope boundary, reported separately and used485in sensitivity analyses, and is treated in Section~\ref{sec:data-core}. Amendment A-005 records the486redefinition, so that the code labels of the protocol, of the implementation and of this manuscript now487agree.488489\subsection{Amendments A-001 and A-002 and their evidence}490491Two eligibility decisions were taken after the protocol freeze, on the evidence of the first run of the492corpus-definition script, and both are logged in \texttt{protocol/amendments.csv} with their date, their493reason and their status (executing agent, pending confirmation by the principal investigator at494checkpoint H-0). They are reported here in full because an eligibility rule changed after preregistration is495exactly the kind of decision a reader must be able to audit \citep{nosek2018preregistration}.496497\emph{A-001, document types.} The protocol's type list was written from the OpenAlex documentation available498at the time; the 2026 snapshot uses the types \texttt{conference-paper} and \texttt{software-paper}, which the499list did not contain. The first run of the eligibility stage excluded under code E2 a very large number of records typed500\texttt{conference-paper}, close to half of them DBLP-linked; the figures of that superseded run are recorded in501the amendment entry and are not repeated here, since the run was not retained. Conference papers are the principal publication form of several areas of502computer science, and their exclusion would have invalidated the composition analyses of Part~IX. The503amendment adds the two types to the list and, more consequentially, makes DBLP linkage decisive: a504DBLP-linked record is a computer-science research output whatever OpenAlex calls it, unless its type is one505that can never be a research output. The same amendment keeps DBLP-linked XPAC records. The corrected run506yields the counts reported above, in which \PrismaEligibleByTypeConferencePaper{} eligible works are507conference papers.508509\emph{A-002, generic deposit platforms and date precision.} Inspection of the eligible corpus by source510(\texttt{systematic\_review/inspect\_sources.py}) showed that a single generic repository, Zenodo, contributed511several times more eligible records dated 2026 than dated 2025, that almost none of them carried a reference512list, and that random samples were slide decks, datasets, software releases and project reports typed as513articles because the depositor had chosen that type; the counts are recorded in the amendment entry. Left in the514corpus, such records would have produced a spurious surge of 2026 output concentrated in a repository that515does not publish research articles, and would have done so in the months where right truncation already516makes the series fragile. The amendment excludes records whose primary source is a generic deposit platform517unless DBLP or arXiv vouch for them, and introduces the \texttt{date\_precision} flag discussed in518Section~\ref{sec:data-dates}. The rule is a source-level rule, not a content judgement: a genuine research519report deposited only on Zenodo is lost, which is accepted because the alternative, keeping every deposit,520contaminates the primary estimand of RQ1. The sensitivity analyses of Part~IX report the monthly series with521and without the generic-deposit exclusion.522523The amendments illustrate a general point about preregistered bibliometrics. The protocol can freeze the524\emph{rules} but not the data model of a database that changes between the time of writing and the time of525extraction; when the data model changes, the honest course is an amendment with its evidence, not a silent526adaptation.527528\section{The computer-science core, the penumbra and the strict core}529\label{sec:data-core}530531The decision of what counts as computer science is the single most consequential modelling choice in a532discipline-level mapping, and the protocol makes it with a three-part membership rule (decision D4) rather533than with any one source. A record that survives the exclusions E1--E7 belongs to the \emph{core} if at534least one of the following holds:535536\begin{enumerate}537\item it is linked to a DBLP record (venue-based membership);538\item its OpenAlex primary field is Computer Science (classifier-based membership);539\item it is linked to an arXiv e-print whose primary category is in the \texttt{cs} archive540  (author-declared membership).541\end{enumerate}542543The third condition is the form given by amendment A-006. As frozen, it required in addition that the record544carry a computer-science topic in OpenAlex, which made the clause unreachable for exactly the works the second545pass recovers (works with no computer-science topic that an arXiv e-print declares as computer science); the546amendment removes the extra requirement and makes the corpus-definition script assert that every record547surviving the exclusions falls into exactly one of three classes, core, penumbra or the residual class548counted below. A record that survives the exclusions, carries a549computer-science topic and satisfies none of the three conditions, which means it has neither venue nor550author support, belongs to the \emph{penumbra}. The three conditions correspond to three independent551judgements about disciplinary membership: the judgement of DBLP's editors about venues, the judgement of552OpenAlex's classifier about content, and the judgement of the authors who chose a primary arXiv category.553Requiring any one of them removes the over-inclusion that a secondary topic alone would produce, while keeping554works that any single source would have missed.555556\subsection{Counts}557558After the sequential exclusions, \PrismaAfterExclusions{} loaded records remain. Of these,559\PrismaCsCoreEligible{} satisfy the membership rule and constitute the eligible mapping corpus,560\PrismaCsPenumbraEligibleOtherwise{} form the penumbra, and \PrismaNeitherCoreNorPenumbraAfterExclusions{}561belong to neither: they carry no computer-science topic, having been loaded by the second pass, and are562neither DBLP-linked nor linked to an e-print with a \texttt{cs} primary category. By construction of the563second-pass identifier list, such records can only have been reached through the journal DOI of an e-print564harvested in the \texttt{cs} set but whose primary category lies in another archive (a cross-listed565e-print), and they are outside the corpus by the same reasoning that excludes the penumbra; this reading566follows from the construction and has not been verified record by record, which the linkage review will do.567The three classes partition the survivors exactly, and the corpus-definition script stops if they do not.568The penumbra is thus a substantial fraction of the569records that a naive ``any computer-science topic'' filter would have called computer science; its exclusion570from the primary analyses is the main difference between the present corpus and what an OpenAlex query alone571would return, and it is the reason why no count in this manuscript should be compared directly with the counts572that the OpenAlex interface displays for field~17. Within the eligible corpus, the three membership573conditions overlap: \PrismaEligibleMembershipInDblp{} works are DBLP-linked,574\PrismaEligibleMembershipCsPrimary{} have a computer-science primary field, and575\PrismaEligibleMembershipArxivLinked{} are linked to an arXiv e-print; \PrismaEligibleMembershipInAcl{} are576ACL Anthology papers. The eligible corpus divides by period into \PrismaEligibleByPeriodBaseline{} works whose577first date falls in the baseline window, \PrismaEligibleByPeriodPost{} in the post-boundary window, and a578small residue of \PrismaEligibleByPeriodPre{} whose first date, taken from a preprint or an e-print, precedes5792018 although the version of record falls in the window; the residue is kept in the corpus and assigned to580the ``pre'' period, so that it enters no monthly series. Figure~\ref{fig:prisma} renders these counts as a581PRISMA-style flow, drawn from the same JSON file that generates the macros of this section582\citep[cf.][]{haddaway2022prismaflow}.583584\begin{figure}[htbp]585\centering586\includegraphics[width=0.9\textwidth]{fig_prisma_flow.pdf}587\caption{PRISMA-style flow of record identification, exclusion and the final mapping corpus. Identification588counts are per source and overlap; exclusions are sequential in the order E1--E7; the core, the penumbra and589the small residual class of second-pass records with neither venue nor author support (text) partition the590records that survive the exclusions. Source: \texttt{data/metadata/prisma\_counts.json},591rendered by \texttt{figures/make\_figures.py}.}592\label{fig:prisma}593\end{figure}594595\subsection{The strict core}596597The membership rule is a disjunction, and a sceptical reader may object that a disjunction of three imperfect598judgements inherits the errors of each. The protocol anticipates the objection with a \emph{strict core},599defined as the conjunction of the first two conditions: works that are both DBLP-linked and labelled computer600science by the OpenAlex classifier. The strict core contains \PrismaStrictCoreDblpAndCsPrimary{} works. Every601headline result of Part~IX is re-estimated on the strict core, and the manuscript reports a result as robust602to scope only when the sign and the order of magnitude agree between the eligible corpus and the strict603core. The strict core is not a better corpus; it is a narrower one, biased toward established venues and604against preprints, dissertations and the newer conference series that DBLP has not yet adopted, and its role605is to bound the influence of scope decisions, not to replace them.606607\subsection{What the design does not do}608609Three limits deserve statement. First, DBLP membership is venue membership: a paper in a DBLP-indexed journal610of bioinformatics or operations research is in the core even when its content is not computational, and the611taxonomy of Part~IV, not the scope rule, is what distinguishes such works. Second, the rule has no language612component; works in any language are eligible, and the share of non-English records, which is concentrated613in dissertations and in regional journals, is reported among the quality indicators. Third, the rule is614applied to the records loaded by the two snapshot passes; a computer-science work that OpenAlex does not615index at all, or indexes without any computer-science topic and without a DBLP or arXiv DOI, is not observed.616The DBLP-only table and the unmatched ACL entries bound the size of this blind spot for two sources with617complete venue lists, and Section~\ref{sec:data-quality} reports them.618619\section{Dates, date precision, indexing lag and right truncation}620\label{sec:data-dates}621622Monthly series are the primary temporal unit of this study, and the November 2022 boundary falls inside a623year. Everything in the structural-change analysis of Part~IX therefore depends on the quality of dates at624the month level, which is considerably worse than at the year level in every bibliographic database. This625section documents three distinct problems, their diagnosis in the corpus, and the rule adopted for each.626627\subsection{Which date}628629Each eligible work carries up to three dates: the OpenAlex publication date of the surviving record, the630publication date of its merged preprint record where one exists, and the arXiv first-submission date where the631work is linked to an e-print. The \emph{first date} of a work is the earliest of the three632(Section~\ref{sec:data-dedup}), except that a publication date of the form \emph{YYYY}-01-01, which is in633most cases a year-only date, is set aside whenever a preprint or e-print date exists, so that the first date634is then the earlier of those two; the \emph{first month} is its calendar month. The first date measures635the entry of the work into the public record, which is the quantity relevant to questions about the pace and636composition of research activity; the publication date of the version of record measures the pace of the637publishing system and is kept for the analyses of venues and of preprint-to-publication lags. The two638coincide for works without a preprint, which are the majority outside the machine-learning areas.639640\subsection{Date precision and the January artefact}641642A large share of publication dates in OpenAlex are year-only dates stored as the first of January. The643phenomenon is visible without any model: the raw monthly series of first appearances shows a spike in every644January that no publishing calendar could explain (Figure~\ref{fig:volume-monthly-raw}), and inspection of the645eligible corpus by year found that between about a fifth and somewhat over a quarter of records carry a date of the form646\emph{YYYY}-01-01, with the proportion varying by source and type; dissertations, book chapters and records647from repositories are the most affected, journal articles with publisher-registered issue dates the least.648Amendment A-002 introduced the flag \texttt{date\_precision}, equal to \texttt{year} when the publication649date is the first of January and the work has neither a merged preprint date nor an arXiv first-submission650date, and \texttt{month} otherwise. A work with a 1 January publication date and an e-print is thus651month-precision, because its first date is taken from the e-print rather than from the publication date652(Section~\ref{sec:data-dedup}). The rule is a heuristic with a known cost: a work genuinely published on6531 January and without a preprint is classed as year-precision. In the eligible corpus,654\PrismaEligibleDatePrecisionMonth{} works are month-precision and \PrismaEligibleDatePrecisionYear{} are655year-precision.656657All monthly series in Part~IX use month-precision records only. Year-precision records enter the annual series,658where their date is correct, and the composition analyses that do not depend on the month. The exclusion of659year-precision records from monthly series is not innocuous: if the share of year-only dates differs between660categories or between periods, the monthly composition is biased relative to the annual one. Part~IX661therefore reports, for each category, the share of year-precision records by year, and treats a divergence662between the monthly and the annual growth estimates as a warning rather than as two independent results.663664\begin{figure}[htbp]665\centering666\includegraphics[width=0.95\textwidth]{fig_volume_monthly_raw.pdf}667\caption{Monthly first appearances in the eligible corpus before the precision filter. The January spikes668reflect year-only publication dates stored as 1 January and motivate the month-precision rule. Source:669\texttt{results/bibliometrics/volume\_monthly.csv}.}670\label{fig:volume-monthly-raw}671\end{figure}672673\subsection{Indexing lag and why \texttt{created\_date} cannot measure it}674675A work is observable only once the database has indexed it, and the delay between publication and indexing676determines how incomplete the most recent months are. The research plan proposed to estimate the delay677profile from the OpenAlex \texttt{created\_date} field, the date on which the record was created in OpenAlex,678and to use the profile to correct the right tail of the monthly series. The diagnostic run of this estimator679(\texttt{results/bibliometrics/indexing\_lag.csv}) refuted the proposal. For works published in 2018, the680median difference between \texttt{created\_date} and the publication date is close to eight years, and the681median falls by almost exactly one month for each later publication month, which is the signature of a682constant creation date: the field records the date on which the record was created in the current OpenAlex683data model, apparently during a re-ingestion in the second half of 2025, not the date on which the work was684first indexed. For the great majority of works published before that re-ingestion, \texttt{created\_date}685carries no information about indexing delay, and for works published after it the field measures the delay686only for the records created since. A lag profile estimated from it would be an artefact of database687maintenance. The field is therefore not used, and the plan's lag-corrected estimates are replaced by the688approach below. The episode is recorded in detail because a reader who reproduces the estimator on a later689snapshot may find the field meaningful again, and because it shows why the mandate's requirement to preserve690immutable raw snapshots is not a formality: the only way to measure indexing lag honestly is to compare two691snapshots of the same database.692693\subsection{Right truncation}694695The last months of the window are incomplete for three distinct reasons that must not be conflated: works696published but not yet indexed (indexing lag), works indexed but not yet linked to a DBLP record because DBLP's697own curation lags the publishers, and works whose citation counts have had little time to accumulate. The698first two shrink the counts of the recent months; the third affects only impact indicators and is treated in699Part~IV through fixed citation windows, which are the standard remedy for the age dependence of citation700counts \citep{wang2013citationwindow,waltman2016review}.701702For the counts, the diagnosis uses the sources against one another. The DBLP dump of 2026-11-02 holds, for7032026, a count well below its 2025 count although ten months of the year have elapsed, which shows that DBLP's704own coverage of 2026 is incomplete; and the eligible corpus for 2026 is of the same order as that for 2025705(\PrismaEligibleByYearTwoZeroTwoSix{} against \PrismaEligibleByYearTwoZeroTwoFive{}), which on its face706suggests that OpenAlex is \emph{ahead} rather than behind. The two observations are reconciled by the707composition of the 2026 records: a much lower share with a reference list, a much lower mean number of708references, a higher share of preprints and, before amendment A-002, the Zenodo contribution described709above. In other words, the recent months are not simply truncated; they are differently composed, with the710fast channels (preprint servers, repositories) over-represented relative to the slow ones (journals,711curated indexes). A correction factor estimated from totals would be wrong for every category whose channel712mix differs from the average.713714The rule adopted is conservative. The ``complete'' observation window for monthly inference ends in715June~2026; the months from July to November 2026 are reported descriptively, separately, and never enter a716breakpoint test. Within the complete window, the structural-change models include the last months with717their observed values and the robustness section of Part~IX re-estimates every break with the window ending718in March 2026 and in August 2026, so that a result that depends on the last months of the window, in either719direction, is identified as such. The API increment720harvested after the snapshot (Section~\ref{sec:data-sources}) adds the works registered between the snapshot721date and the extraction date, which narrows but does not close the gap. Finally, forward-dated records,722works carrying a 2026 publication date for an issue not yet released, are retained when their date precedes723the extraction date because they are, by the definition of the first date, in the public record; the724current build does not count them separately, and a count by month of the records whose publication date lies725after the snapshot date is listed among the quality indicators to be added before Part~IX is finalised.726727\section{Quality indicators and known biases}728\label{sec:data-quality}729730PRISMA 2020 asks for an assessment of the risk of bias in the included studies; a mapping study has no731studies in that sense, and the protocol replaces the item with indicators of the quality of the bibliographic732record itself, reported per year and per category in Appendix~B. This section defines the indicators,733reports the corpus-level values that are available at the time of writing, and sets out the known biases of734the design, each with the direction in which it is expected to push the results of Part~IX.735736\subsection{Metadata completeness}737738Four completeness indicators are computed for every work: presence of an abstract, presence of a non-empty739reference list, presence of at least one affiliation with an institution identifier, and presence of a740country for at least one author. In the eligible corpus, \PrismaEligibleMembershipHasAbstract{} works have741an abstract and \PrismaEligibleMembershipHasReferences{} have at least one resolved reference, out of742\PrismaCsCoreEligible{}. Both shares are lower than those reported for the DOI-bearing journal literature in743comparisons of OpenAlex with the commercial indexes \citep{culbert2025openalexcoverage,delgadoquiros2024completeness},744which is expected in a corpus that includes dissertations, book chapters and preprints, and both fall sharply745for the most recent year (Figure~\ref{fig:types-openness}). The consequences are concrete. A work without an746abstract cannot be embedded or classified by its content and is classified from its venue, its arXiv747categories and its OpenAlex topics alone, with a lower confidence score; a work without references does not748contribute out-citations to the knowledge graph, so that the citation flows of Part~VIII are estimated on the749subset with references and are weighted accordingly. The ACL Anthology abstracts fill part of the first gap750for one subfield; no source fills the second.751752\begin{figure}[htbp]753\centering754\includegraphics[width=0.95\textwidth]{fig_types_openness.pdf}755\caption{Document-type composition and metadata completeness of the eligible corpus by publication year756(2026 incomplete). Source: \texttt{results/bibliometrics/volume\_yearly.csv}.}757\label{fig:types-openness}758\end{figure}759760\subsection{Coverage overlap between sources}761762The overlap between sources is the study's main instrument for measuring its own coverage, because each763source has a complete list of what it holds. Three overlaps are reported. The DBLP--OpenAlex overlap, after764both passes, is the share of DBLP records in the window with an OpenAlex counterpart; its complement, the765DBLP-only table, is tabulated by year and type in Appendix~B. The arXiv--OpenAlex overlap is the share of766\texttt{cs} e-prints with an OpenAlex counterpart (\LinkArxivLinkedDistinct{} of767\LinkArxivCreatedTwoZeroOneEightTwoZeroTwoSix{} e-prints first submitted in the window). The ACL--OpenAlex768overlap is the share of Anthology papers linked (\LinkAclLinkedDistinctTwoZeroOneEightTwoZeroTwoSix{} of769\LinkAclTwoZeroOneEightTwoZeroTwoSix{}). An unmatched ACL paper is a documented computer-science publication770that the backbone either does not hold or holds under a title or identifier the linkage could not resolve;771the residual is therefore an upper bound on the sum of backbone non-coverage and linkage failure for one772conference-heavy subfield, and a manual review of a sample of unmatched entries, planned with checkpoint773H-1, separates the two.774775Figure~\ref{fig:volume-yearly} shows the eligible works by publication year and by source membership. The776widening of the arXiv-linked layer and the relative stability of the DBLP-linked layer are the first777descriptive facts of the study, and they are also a warning about right truncation: in the most recent778years, the fastest source dominates.779780\begin{figure}[htbp]781\centering782\includegraphics[width=0.95\textwidth]{fig_volume_yearly.pdf}783\caption{Eligible works per publication year by source membership (DBLP-linked, OpenAlex computer-science784primary field, arXiv-linked, ACL Anthology); memberships overlap. 2026 is incomplete. Source:785\texttt{results/bibliometrics/volume\_yearly.csv}.}786\label{fig:volume-yearly}787\end{figure}788789\subsection{Known biases and their expected direction}790791\emph{DBLP coverage bias.} DBLP indexes established venues and adopts new conference series with a delay;792within the core, the DBLP-linked subset over-represents long-standing journals and conferences and793under-represents workshops, regional venues and the newest series. Because the strict core is defined794through DBLP, every strict-core estimate inherits this bias, which runs against emerging topics and against795the most recent years. Results that are stronger in the strict core than in the eligible corpus are therefore796unlikely to be artefacts of over-inclusion; results that appear only in the eligible corpus may be.797798\emph{OpenAlex topic noise.} The topic classifier mislabels in both directions, and the membership rule799bounds only the over-inclusion, through the penumbra. Under-inclusion is bounded by DBLP and arXiv for the800venues and preprints they cover and is unbounded elsewhere. Within the core, the topic labels are used as801one of several signals in the taxonomy, never alone, and the taxonomy's validation on human-reviewed samples802(Part~IV, checkpoint H-2) is the only direct measure of how much label noise survives.803804\emph{XPAC records.} The expansion records excluded under E3 are not a random sample of the literature: they805are concentrated in repositories, in non-English regions and in recent years. Their exclusion removes some806genuine research, mostly outside the venue system, and biases the corpus toward the formal literature; the807DBLP exception restores the part that the venue system vouches for. The sensitivity analysis that reinstates808XPAC records of research types reports the difference.809810\emph{Generic-deposit platforms.} The exclusion under E7 removes, besides the deposits that motivated it, any811research output whose only home is a generic repository; the effect is largest in 2025--2026 and in the812areas, such as scientific computing and computational biology, where data and software deposits accompany813papers. The direction of the bias is toward under-counting recent output in those areas, which is the814opposite of the bias the rule corrects.815816\emph{Missing abstracts.} Abstract availability depends on publishers' metadata deposits and on OpenAlex's817sources; it is lower for conference papers of some societies and for older records, and it is strongly818correlated with the DOI registrant. Every content-based analysis is conditional on abstract availability, and819the appendix reports the abstract share by category and year so that the reader can judge where the820conditional population departs from the corpus.821822\emph{Forward-dated and year-dated records.} Both displace works along the time axis in a direction that is823not symmetric: year-only dates pull works toward January, forward dates push them toward the future. The824month-precision rule removes the first; the second is bounded by the window end and will be counted825explicitly.826827\emph{Retractions.} Retracted works are in the corpus and out of the impact analyses. The retraction flag828depends on the coverage of the Retraction Watch data through Crossref, which is incomplete for publishers829that do not deposit retraction notices; some retracted works are therefore in the impact analyses830unflagged. The direction of the bias is small but known: retraction is concentrated in a few venues and in831a few years, and the landmark selection of Part~V checks each candidate against the current retraction832status at the time of review.833834\emph{Preprint duplication.} The residual double count after the exact-title pairing inflates the areas that835preprint most, which are also the areas expected to show the strongest growth. The sensitivity with a836similarity threshold and the removal of preprints with a journal DOI bound the effect; a growth result for837machine-learning areas that does not survive these two variants is reported as not robust.838839None of these biases is corrected by reweighting. The study reports them, bounds them where a second source840permits, and re-estimates its results on alternative corpora (strict core; corpus with penumbra; corpus with841XPAC research types; corpus with DBLP-only records; corpus with generic deposits) so that the reader can see842how far each conclusion depends on each scope decision. This is the approach that the science-of-science843literature recommends for databases whose coverage is known to be uneven \citep{fortunato2018scienceofscience}, and it844is the only one compatible with a mandate that forbids the fabrication of corrections.845846\section{Research-integrity procedures}847\label{sec:data-integrity}848849The mandate for this working paper lists integrity rules that are unusual for a bibliometric study: no850fabricated reference, identifier, count or result; no paper content invented from metadata; no abstract851treated as a full text; no model-generated summary treated as evidence; no temporal association read as852causation; no preprint described as peer-reviewed; inaccessible data recorded explicitly; and human review of853every scientific conclusion. The rules are a response to the way the study is conducted, by a principal854investigator directing autonomous software agents, and this section describes the procedures by which they855are enforced. The procedures are reported here, in the data part, because most of them operate on data and856artefacts rather than on prose.857858\subsection{Preregistration and the amendments log}859860The protocol (\texttt{protocol/PROTOCOL.md}) was frozen on 2026-11-02 before any analysis of the corpus. It861fixes the research questions, eight directional hypotheses H1--H8 with the sign of the expected effect862stated so that null and opposite results must be reported as such, the eligibility criteria and their codes,863the sources and extraction dates, the data items, the quality indicators, the synthesis methods, the864human-review checkpoints and the data-availability commitments. The preregistration is internal, in the865repository rather than in a registry, which limits its evidential value relative to a time-stamped public866record \citep{nosek2018preregistration}; the repository's version history and the dated amendments log are867the available substitute, and the release of Phase~9 will include both. Every departure from the frozen868protocol is an entry in \texttt{protocol/amendments.csv} with an identifier, a date, the phase, the change,869the reason and the approver, which until the principal investigator signs is ``executing agent (pending870PI)''. At the time of writing the log holds seven entries: A-000 records that the mandate's acceptance871criteria were completed by the executing agent after the pasted text was truncated; A-001 and A-002 are the872eligibility amendments of Section~\ref{sec:data-eligibility}; A-003 and A-004 fix the budgets and the models873of the LLM-based taxonomy validation and knowledge-graph extraction; A-005 records the renumbering of code874E7 and A-006 the revised arXiv clause of the membership rule (Sections~\ref{sec:data-eligibility}875and~\ref{sec:data-core}), both prompted by the writing and critique passes of the first manuscript draft. Analyses876added after the freeze are labelled exploratory in Part~IX.877878\subsection{Immutable raw snapshots and provenance}879880Nothing in the corpus can be traced to a result unless the raw data from which it was built can be881re-obtained or re-verified. The OpenAlex release is a public, versioned object; the study keeps its manifest,882the per-file content lengths and SHA-256 hashes, and the filtered subset, rather than a copy of the full883bucket (decision D6), so that a reader with the same release and the same scripts obtains byte-identical884inputs, and a reader with a later release can measure the difference. The DBLP dump is kept with its MD5,885the arXiv harvest as the raw OAI-PMH pages, the ACL export with its SHA-256, and the API increment as the raw886JSON pages with a ledger of calls and cost. Each ingestion job writes a ledger with attempts, timings, hashes,887counts and errors, which is the job-level provenance the mandate requires. The processing chain from raw888snapshot to PRISMA counts is a sequence of scripts with measured wall times in889\texttt{reproducibility/RUNBOOK.md}, executed in a locked Python environment whose package versions are890recorded; the chain was executed on a single node, deliberately, because it is network-bound and because a891single-node execution is easier to reproduce than a distributed one.892893\subsection{Numbers that cannot be typed}894895Every count, share and estimate quoted in this manuscript is a LaTeX macro generated by896\texttt{tables/make\_numbers.py} from the result files (\texttt{prisma\_counts.json},897\texttt{linkage\_report.json}, the bibliometric and econometric result tables). The macros cannot be edited898by hand, a macro whose source file is absent is undefined and stops the compilation, and a number the authors899need but the pipeline has not produced appears in the text as a visible pending marker rather than as a900figure. The procedure is the mechanical form of the mandate's requirement that every quantitative claim901trace to a script, an input artefact and a result file. It has a cost in prose, since the text cannot902anticipate a number, and a benefit in auditability, since a reader who regenerates the result files903regenerates the manuscript's numbers.904905\subsection{Verified references}906907The bibliography of this manuscript is built under a rule that no reference may be cited unless it has been908resolved to a bibliographic record by a single programmatic lookup against Crossref, OpenAlex or arXiv, with909the returned title compared with the intended one and the comparison recorded in a verification table910(\texttt{references\_verification*.csv}: key, identifier, service, returned title, match). References that911fail the lookup are removed from the bibliography and kept in the table with a failed status, and a second912agent re-verifies a sample in the audit phase. The rule exists because the model-based agents that draft the913prose are known to produce plausible but non-existent references when unconstrained; it is also why914preprints are cited as preprints, with their arXiv identifier and without a venue, and why the bibliography915distinguishes a technical report from a peer-reviewed article in its entry type.916917\subsection{Agents, artefacts and the register}918919The study is executed by a research-director agent that spawns specialised agents with written briefs;920each brief, the artefacts it produced and the agent that verified them are recorded in921\texttt{agents/REGISTER.md}. The rule that governs the register is that no agent approves its own findings:922an artefact enters the manuscript only after a different agent, the scientific critic or the reproducibility923auditor, has checked it, and the register records the checker. Agents exchange structured artefacts (tables,924JSON, Markdown with declared schema) rather than conversational summaries, so that what one agent passes to925another can be inspected after the fact. The writing agents, including the one that produced this part, work926from the source-of-truth files named in their briefs and are instructed to describe what the code does927rather than what the plan intended; where the two differ, as with the indexing-lag estimator of928Section~\ref{sec:data-dates}, the difference is reported.929930\subsection{Human-review checkpoints}931932Six checkpoints are defined (H-0 to H-5: plan and protocol; linkage sample; taxonomy validation sample;933landmark list and specialist reviews; structural-change results and their interpretation; manuscript934conclusions), each with a named artefact that the principal investigator reviews and signs935(\texttt{reports/HUMAN\_REVIEW\_CHECKPOINTS.md}). Until a checkpoint is signed, every conclusion that936depends on it is labelled ``pending human review'' in the text, with the same visible marker as a missing937number. At the time of writing no checkpoint is signed. The reader should take this literally: the938conclusions of this manuscript are, in its present version, the conclusions of a procedure, reviewed by its939own audit agents and awaiting the judgement of its human author.940941\subsection{What the procedures do not guarantee}942943The procedures guarantee traceability, not truth. A count traced to a script is still wrong if the script is944wrong; a verified reference is still miscited if the prose misreports what it says; a preregistered945hypothesis can still be tested with an inappropriate model. What the procedures make possible is detection:946every number, reference and decision has a place where an auditor can look. The remaining risks are the947ones this part has tried to state plainly: the coverage and labelling errors of the sources, the residual948duplication, the fragility of the recent months, and the dependence of every scope decision on the949judgement, recorded but not infallible, of the agents that took it.950951