{ "title": "Rich textual-data sources: the catalogue", "dates": { "profiles": "2026-10-11", "provider_checks": "2026-10-29" }, "scope": "Representative review of provider documentation and metadata. No source datasets acquired, proposed tasks executed or study fit validated.", "narrative_exclusion": "D17", "groups": [ { "id": "institutional", "name": "Institutional records and organizational claims", "arenas": [ "A01", "A02", "A03" ] }, { "id": "products", "name": "Products, innovations and evaluations", "arenas": [ "A04", "A05", "A06", "A07", "A08" ] }, { "id": "interaction", "name": "Interaction and process records", "arenas": [ "A09", "A10", "A11", "A12" ] }, { "id": "structured", "name": "Complementary structured data", "arenas": [] } ], "arenas": [ { "id": "A01", "name": "Regulation and public administration", "group": "institutional", "families": [ "D10" ], "concrete_sources": "Federal Register", "fit_and_boundary": "Framing and stated obligations; text is not implementation or behavior." }, { "id": "A02", "name": "Corporate disclosure", "group": "institutional", "families": [ "D09" ], "concrete_sources": "SEC EDGAR", "fit_and_boundary": "Issuer claims and disclosed events; disclosure and decision dates differ." }, { "id": "A03", "name": "Work and occupations", "group": "institutional", "families": [ "D22" ], "concrete_sources": "O*NET; job postings", "fit_and_boundary": "Task descriptions and instrument design; no direct evidence of automation or authority." }, { "id": "A04", "name": "Product launches and descriptions", "group": "products", "families": [], "concrete_sources": "Product Hunt", "fit_and_boundary": "Described functionality; posts, containers and companies are different units." }, { "id": "A05", "name": "Patents", "group": "products", "families": [ "D08" ], "concrete_sources": "USPTO PatentsView", "fit_and_boundary": "Technological comparison sets; patents do not cover all innovation." }, { "id": "A06", "name": "Scientific text and knowledge links", "group": "products", "families": [ "D11", "D16" ], "concrete_sources": "OpenAlex; QASPER", "fit_and_boundary": "Abstract/reference analysis and evidence-grounding fixtures; incomplete fields and full-text rights." }, { "id": "A07", "name": "Product composition and artifacts", "group": "products", "families": [ "D03", "D04", "D07" ], "concrete_sources": "Open Food Facts; Food.com; RecipePairs", "fit_and_boundary": "Attribute/recombination comparisons; snapshot novelty is not historical firstness." }, { "id": "A08", "name": "Reviews and responses", "group": "products", "families": [ "D01", "D02", "D05" ], "concrete_sources": "Amazon Reviews 2023; Google Local; Goodreads", "fit_and_boundary": "Evaluative categories and response sequences; selection and reuse limits remain." }, { "id": "A09", "name": "Meetings and interview accounts", "group": "interaction", "families": [ "D12" ], "concrete_sources": "AMI; publisher podcast transcripts", "fit_and_boundary": "Situated discussion or mediated accounts; source types are not interchangeable." }, { "id": "A10", "name": "Collaborative discussion and platforms", "group": "interaction", "families": [ "D13" ], "concrete_sources": "WikiConv; Hugging Face", "fit_and_boundary": "Contribution and ecosystem histories; current observations are not complete history." }, { "id": "A11", "name": "Human–AI writing and creative traces", "group": "interaction", "families": [ "D14", "D19", "D20", "D25" ], "concrete_sources": "CoAuthor; Noy–Zhang; Doshi–Hauser; WildChat annotations", "fit_and_boundary": "Writing/generation sequences and comparisons; private criteria and stopping reasons remain unobserved." }, { "id": "A12", "name": "Text on four platforms", "group": "interaction", "families": [], "concrete_sources": "", "fit_and_boundary": "Descriptions, comments, captions and transcripts differ." } ], "families": [ { "id": "D01", "name": "Amazon Reviews 2023", "checked_at": "2026-10-11", "group": "products", "arena": "A08", "kind": "arena", "provider": "McAuley Lab", "release": "2023 release; creator Hugging Face metadata revision 2b6d039ed471f2ba5fd2acb718bf33b0a7e5598e (modified 2024-12-08). This pins the repository metadata, not every externally hosted category file.", "units_and_scope": "User–item reviews and product metadata, across Amazon categories.", "size": "Creator reports 571.54 million reviews, 54.51 million users and 48.19 million items; metadata coverage is not identical to reviewed-item coverage.", "coverage_and_dates": "Review dates May 1996–September 2023. Product metadata is a snapshot, not a historical price or product-attribute panel.", "fields_and_links": "Review text/title, rating, user_id, timestamps, verified purchase, helpful votes and images; item title, descriptions, features, categories, store and price. Join reviews to metadata with parent_asin; do not silently reuse older ASIN join assumptions.", "access": "public-download; reuse unclear", "route": "Creator category JSON downloads and Hugging Face repository.", "authentication": "Creator links and HF metadata are public; HF gated=false. File acquisition was not tested.", "inspected": "Primary documentation and native HF metadata inspected. Do not execute the custom HF loader; use native data files in a later acquisition.", "reuse_status": "unknown", "reuse_terms": "No explicit dataset license found on the inspected creator site or HF metadata. Public availability does not establish permission to redistribute review text or images.", "reuse_group": "unresolved", "limitations": [ "Ratings and helpful votes are reception measures, not novelty labels or an unbiased quality standard.", "Items are not firms; review selection and product survival can distort organizational inferences.", "Actual files need schema inspection: documentation and examples vary in timestamp/helpful-vote naming and price type. User identifiers and review text need a deliberate sharing plan." ], "suggested_task": "Separate claims about product attributes from customer experience and unsupported inferences; compare conventional text features with constrained LLM extraction." }, { "id": "D02", "name": "Google Local Data 2021", "checked_at": "2026-10-11", "group": "products", "arena": "A08", "kind": "arena", "provider": "McAuley Lab", "release": "US Google Local 2021 collection, through September 2021; no per-file checksum manifest inspected.", "units_and_scope": "Dated reviews of local businesses, business responses, and business metadata.", "size": "666,324,103 reviews; 113,643,107 users; 4,963,111 businesses. Vermont partition: 852,203 reviews and 11,291 businesses.", "coverage_and_dates": "Review time and response time are available. Business attributes are snapshot values; collection coverage is not an operating-history panel.", "fields_and_links": "Review user_id, time, rating, text, pictures, resp.time and resp.text; business gmap_id, name, address, categories, coordinates, hours, price and related-business identifiers. Join on gmap_id.", "access": "public-download; reuse unclear", "route": "Creator per-state metadata/reviews, 10-core subsets and ratings-only links.", "authentication": "No authentication stated for direct links; no data downloaded.", "inspected": "Creator data format, files and citation inspected.", "reuse_status": "unknown", "reuse_terms": "No explicit dataset reuse license found on the dataset-specific page. Do not infer an open license from UCSD hosting.", "reuse_group": "unresolved", "limitations": [ "Ratings-only files cannot support the proposed text task; 10-core filtering changes the target population.", "A response does not establish that remediation occurred, and a related-business link does not establish common ownership.", "Even a state partition is large. Business names and user information should not be copied into a public sample without checking permitted reuse." ], "suggested_task": "Construct evidence-linked complaint, promised action and response-lag observations, explicitly retaining missing responses." }, { "id": "D03", "name": "Food.com Recipes and User Interactions", "checked_at": "2026-10-11", "group": "products", "arena": "A07", "kind": "arena", "provider": "Kaggle", "release": "Creator Kaggle dataset version 2, updated 2019-11-08T01:18:21.147Z; native metadata inspected.", "units_and_scope": "Recipes and user–recipe interactions from the 2000–2018 collection.", "size": "231,637 recipes, 226,570 users and 1,132,367 reviews in the UCSD directory; Kaggle package metadata reports 280,350,373 bytes. The paper's filtered analysis sample is smaller.", "coverage_and_dates": "Recipe submitted dates and interaction dates allow a declared temporal reference set; exact date completeness remains untested.", "fields_and_links": "Recipes: id, name, contributor_id, submitted, ingredients, steps, tags, nutrition, minutes and counts. Interactions: user_id, recipe_id, date, rating and review. Link recipe_id to recipe id.", "access": "public listing; download authentication unverified; reuse restricted/unclear", "route": "Original shuyangli94 Kaggle listing; do not substitute a similarly named scrape.", "authentication": "Kaggle download may require an account; no download or login was attempted.", "inspected": "Creator directory, paper, and native Kaggle metadata match this dataset identity.", "reuse_status": "copyright notice; no broad reuse grant found", "reuse_terms": "Creator metadata says “Data files © Original Authors.” A different Food.com scrape's CC0 label cannot license this release.", "reuse_group": "restricted", "limitations": [ "Ratings measure reception, not novelty; nutrition fields do not by themselves validate health claims.", "The paper applies recipe/user filtering and chronological splits: do not treat its filtered counts as the raw package.", "Recipe posting dates indicate publication in this source, not invention dates." ], "suggested_task": "Compare ingredient-pair rarity, text distance and blinded judgments of substantive change. Use recipes with similar prose but different composition and the converse." }, { "id": "D04", "name": "RecipePairs / SHARE recipe-editing data", "checked_at": "2026-10-11", "group": "products", "arena": "A07", "kind": "arena", "provider": "Hugging Face", "release": "Creator HF revision fb87cbf30db41c4214e647a6bfac2f1dcb862c3e, modified 2023-03-21; card configuration 1.5.0.", "units_and_scope": "Derived pairs of recipes and a linked recipe-content table. The original SHARE study's selected benchmark is a different manifestation.", "size": "HF card reports 6,908,697 pair examples; pairs.parquet 71,788,476 bytes and recipes.parquet 169,885,896 bytes. The 2022 paper describes about 83K pairs. Selection/release reconciliation is unresolved.", "coverage_and_dates": "Pair-table dates were not verified; historical novelty requires a checked join to original recipe dates.", "fields_and_links": "Recipe-pair identifiers, similarity and dietary tags; linked recipe content in the separate recipes table. Exact parquet schema and key uniqueness require later inspection.", "access": "public-download; declared GPL-3.0 with upstream-rights caveat", "route": "Creator HF native parquet files.", "authentication": "gated=false in creator metadata; no file acquisition.", "inspected": "Creator card, repository file metadata and primary SHARE paper inspected.", "reuse_status": "declared GPL-3.0; upstream provenance unresolved", "reuse_terms": "HF labels this dataset GPL-3.0. That declaration does not settle rights in the underlying Food.com content or prove the paper's selected benchmark has identical terms.", "reuse_group": "unresolved", "limitations": [ "Pairs are derived/matched examples, not observed human creative lineage, editing decisions or accepted improvements.", "The roughly 83K-paper versus 6.9M-card discrepancy must be resolved before numerical replication.", "Dataset-license labeling alone does not resolve underlying source rights." ], "suggested_task": "Compare lexical distance with changes to ingredients and procedural dependencies. Treat dietary labels as constraints to inspect rather than certified nutritional advice." }, { "id": "D05", "name": "Goodreads / UCSD Book Graph", "checked_at": "2026-10-11", "group": "products", "arena": "A08", "kind": "arena", "provider": "Mengting Wan", "release": "Late-2017 collection with May-2019 revisions documented on the moved-from UCSD Book Graph site. Current acquisition link not verified.", "units_and_scope": "Books, works, authors, series, public user interactions and reviews.", "size": "Creator page reports 2,360,655 books, 1,521,962 works, 829,529 authors, 400,390 series and 876,145 users; revised interactions 229,154,523 and approximately 15.7M reviews. Older counts on the same page describe earlier versions.", "coverage_and_dates": "Public interaction/review dates where present; edition publication and work identity require separate treatment.", "fields_and_links": "Book/user/review IDs; book-to-work, author and series links; shelves, reading/rating actions, text reviews and dates. Genre subsets can overlap.", "access": "academic-use-only; current download unverified", "route": "Creator site moved from the UCSD Google Site to mengtingwan.github.io; old Drive paths are deprecated.", "authentication": "Current download and any account requirement not verified.", "inspected": "Detailed moved-from creator documentation inspected; current page did not render substantive content in the text browser.", "reuse_status": "restricted research reuse", "reuse_terms": "Creator terms limit use to academic research, prohibit redistribution and prohibit commercial use. Do not bundle source reviews in a portable sample.", "reuse_group": "restricted", "limitations": [ "Similar-book graphs may not be closed within a genre subset.", "Reading/rating activity and panel agreement do not measure durable evaluative skill.", "Terms and current download availability make this a less portable starter than explicitly licensed sources." ], "suggested_task": "Compare description-level similarity with audience interpretation while keeping author, work and edition units separate." }, { "id": "D06", "name": "MovieLens 32M", "checked_at": "2026-10-11", "group": "structured", "arena": "", "kind": "structured", "provider": "GroupLens", "release": "Stable ml-32m generated 2023-10-13; per-file MD5 values in the README.", "units_and_scope": "Movie ratings and tag applications by users with at least twenty ratings; movie identity links.", "size": "32,000,204 ratings; 2,000,072 tag applications; 87,585 movies; 200,948 users.", "coverage_and_dates": "Ratings from 1995-01-09 through 2023-10-12; ratings and tags carry timestamps.", "fields_and_links": "movies(movieId,title,genres), ratings(userId,movieId,rating,timestamp), tags(userId,movieId,tag,timestamp), links(movieId,imdbId,tmdbId). No demographics or full review corpus.", "access": "public-download; custom research terms", "route": "GroupLens static ZIP and README.", "authentication": "No login documented for download; no ZIP acquired.", "inspected": "Official release and full README inspected.", "reuse_status": "custom noncommercial/research conditions", "reuse_terms": "Acknowledge GroupLens; redistribution/transformation must retain the applicable conditions; commercial or revenue-bearing use requires permission. This is not an unrestricted open-data license.", "reuse_group": "restricted", "limitations": [ "The at-least-twenty-ratings inclusion criterion selects active users.", "Cross-site IDs do not grant rights to external film content.", "A dated rating records a platform action rather than initial exposure or a selection decision." ], "suggested_task": "Build a transparent linkage demonstration and compare summary ratings with selected content measures." }, { "id": "D07", "name": "Open Food Facts", "checked_at": "2026-10-11", "group": "products", "arena": "A07", "kind": "arena", "provider": "Open Food Facts", "release": "Continuously updated product database; documentation checked 2026-10-11. No dated data snapshot acquired or frozen. Current API documentation identifies v3.6, with v2 deprecated.", "units_and_scope": "Barcode-identified food products with contributed structured fields and label images.", "size": "No current exact product count verified; avoid quoting a moving website total.", "coverage_and_dates": "Record creation/modification timestamps concern database observations, not product launch dates or invention.", "fields_and_links": "Barcode/code joins product names, brands, categories, ingredients, nutrients, labels and images. Brand names are not resolved firm identifiers.", "access": "public access; explicitly licensed database", "route": "Official API and documented bulk export route; bounded API retrieval would suit a later small demo.", "authentication": "Reads do not require authentication; official docs require a custom User-Agent. No data records acquired.", "inspected": "Official API and rights documentation inspected; the main data page did not render in the browser.", "reuse_status": "explicit, component-specific", "reuse_terms": "Database: ODbL; individual database contents: DbCL; images: CC BY-SA. Preserve attribution and applicable share-alike obligations. Third-party label artwork/content can have additional rights; do not assume one license covers every component.", "reuse_group": "explicit", "limitations": [ "Crowdsourced coverage and field missingness can differ systematically by product/region.", "Structured entries are comparison evidence, not infallible ground truth.", "A product snapshot supports relative difference within the reference set; it cannot establish historical novelty or market introduction." ], "suggested_task": "Create a new coded observation of a packaging or ingredient claim, linking each decision to the label source and checking against structured metadata." }, { "id": "D08", "name": "USPTO PatentsView disambiguated patent data", "checked_at": "2026-10-11", "group": "products", "arena": "A05", "kind": "arena", "provider": "U.S. Patent and Trademark Office", "release": "Current USPTO description covers granted and pre-grant releases through 2025-12-31, with annualized derivative panels updated December 2025. Exact ODP file manifest/version was not obtained.", "units_and_scope": "Patents/applications with disambiguated inventor and assignee entities, technology classifications and citations.", "size": "Current row/file counts not verified. Coverage start depends on the selected table; do not transfer an old API count to the current bulk release.", "coverage_and_dates": "Application, publication and grant dates have different meanings. Historical reference sets must use information actually public before each target cutoff.", "fields_and_links": "Patent/application identifiers link bibliographic records, text tables, inventors, assignees, classifications and citations. Disambiguation versions can change longitudinal entity links.", "access": "registration-controlled public source; current release terms unresolved", "route": "USPTO Open Data Portal bulk datasets, including pvgpatdis; legacy PatentSearch is not assumed to be an available route.", "authentication": "ODP registration/API-key requirement announced effective 2026-06-18, with additional profile fields in August 2026. No account or key used.", "inspected": "Current USPTO source and ODP access notices inspected; current file acquisition untested.", "reuse_status": "manifestation-dependent; current terms unverified", "reuse_terms": "A legacy PatentSearch 2024 release explicitly states CC BY 4.0. Do not automatically transfer that statement to every current ODP file; check release-specific terms when acquiring.", "reuse_group": "unresolved", "limitations": [ "Citations can reflect examiners and legal drafting as well as inventor knowledge search.", "Grant timing cannot substitute for information availability; patents are selected inventions rather than the full innovation population.", "Text metrics and citations do not reveal aspiration levels, internal search decisions or commercial value." ], "suggested_task": "Compare technical-component recombination with lexical distance, then construct a small source-linked assignee entry/event table." }, { "id": "D09", "name": "SEC EDGAR filings and company facts", "checked_at": "2026-10-11", "group": "institutional", "arena": "A02", "kind": "arena", "provider": "U.S. Securities and Exchange Commission", "release": "Live submissions and XBRL APIs, updated in real time with nightly bulk packages. A dated source manifest must be frozen for a study; none was acquired here.", "units_and_scope": "Issuer filings, filing events and standardized financial facts; issuer is identified by CIK.", "size": "No finite study population selected. Submission responses include recent history and pointers to older chunks; do not infer full coverage from the initial response.", "coverage_and_dates": "Filing dates, reporting periods and event dates are distinct. XBRL required from 2009; amendments and restatements matter for as-of analyses.", "fields_and_links": "Ten-digit CIK, accession, form, dates and primary document link; names/tickers/exchanges; company facts keyed by taxonomy/concept/unit and filing context. Standardized aggregated facts omit custom tags.", "access": "public access with official dissemination permission", "route": "data.sec.gov submission/companyfacts endpoints and SEC filing documents.", "authentication": "No API key required; identify automated requests and respect the official aggregate limit of ten requests/second.", "inspected": "Official API and privacy/dissemination documentation inspected. No filing records acquired.", "reuse_status": "official copying/distribution permission, with exclusions", "reuse_terms": "SEC policy permits copying/distributing public information with source acknowledgment; logos/artwork have exclusions. Do not label every third-party exhibit public domain merely because EDGAR hosts it.", "reuse_group": "explicit", "limitations": [ "Filings are issuer representations and legally structured disclosures, not direct observation of internal decision processes.", "Current facts and last-filed frames can leak later restatements into historical analyses.", "A financial outcome is not an aspiration threshold, and a disclosure date is not necessarily a decision date." ], "suggested_task": "Create NEW event-level observations: issuer, role, event type, interim status, event date, disclosure date, accession and supporting passage. Resolve duplication and distinguish absence from unreported detail." }, { "id": "D10", "name": "Federal Register documents and metadata", "checked_at": "2026-10-11", "group": "institutional", "arena": "A01", "kind": "arena", "provider": "Office of the Federal Register and Government Publishing Office", "release": "Continuously updated API; published documents from 1994, public-inspection documents from November 2011. Freeze a dated query, identifiers and official source versions for any new dataset.", "units_and_scope": "Regulatory/public-notice documents, their agency associations and separately coded obligations/events.", "size": "No fixed study count obtained. API pagination returns only the first 2,000 results; partition queries by dates rather than assuming that a single broad query is complete.", "coverage_and_dates": "Publication, effective, comment and action dates are different. Date presence and legal status need document-level checking.", "fields_and_links": "Document number, publication date, document type, issuing agencies, title and full-text/PDF references; docket/RIN links where present. Issuing agency is not automatically the implementing actor.", "access": "public access; government-content reuse with exclusions", "route": "Official JSON/CSV API and document pages; use the linked official PDF/govinfo text when legal wording matters.", "authentication": "No API key required; no records downloaded.", "inspected": "Official developer documentation and Library of Congress rights statement inspected.", "reuse_status": "government public-domain component; inspect exceptions", "reuse_terms": "US government content is generally public domain; reprinted third-party copyrighted material can be an exception. NARA/OFR seals are not freely reusable. Do not blanket-license every attachment.", "reuse_group": "explicit", "limitations": [ "The measured outcome is an explicit textual obligation, not observed cooperation or successful implementation.", "Corrections, multiple documents per rule and multiple obligations per document change denominators.", "API XML is informational; official PDF references matter. Existing corpus ownership does not make the new coded dataset already built." ], "suggested_task": "Construct NEW document–obligation records: implementing agency, assigned action, counterpart, authority, date, evidence passage and uncertainty. Distinguish issuing agencies, historical mentions, consultation and explicit implementation duties." }, { "id": "D11", "name": "OpenAlex scholarly graph", "checked_at": "2026-10-11", "group": "products", "arena": "A06", "kind": "arena", "provider": "OpenAlex", "release": "Live graph and public snapshot route; documentation checked 2026-10-11. No current snapshot manifest pinned.", "units_and_scope": "Works, authors, institutions and related scholarly entities linked through identifiers, authorship and citations.", "size": "Current exact graph size not verified; the 2022 paper's size is historical and is not a current release count.", "coverage_and_dates": "Publication dates and snapshot dates differ; entity resolution, affiliations and classifications can change.", "fields_and_links": "Work IDs/DOIs, titles, dates, authorship/institutions, references/citations and abstracts where available. Full article text is not guaranteed by metadata availability.", "access": "CC0 metadata; API account/key for regular use", "route": "Current API access documentation and public snapshot distribution.", "authentication": "Maintainer announced keys for regular API use from 2026-02-13 with limited keyless access. Current quotas/prices require checking at acquisition; no account used.", "inspected": "Creator paper, official blog and dated maintainer access notice inspected; some help pages rendered only shells.", "reuse_status": "CC0 metadata", "reuse_terms": "Creator confirms CC0 underlying data and public snapshots. This does not license separately obtained article PDFs or all full text.", "reuse_group": "explicit", "limitations": [ "Coauthorship and affiliation do not directly measure coordination or decision authority.", "Incomplete abstracts/references and changing disambiguation can shift novelty measures.", "A literature graph is neither a verified effect-size database nor evidence that an agent configuration behaves as advertised." ], "suggested_task": "Build a dated source-linked entry/recombination measure, explicitly tracing institution and topic definitions." }, { "id": "D12", "name": "AMI Meeting Corpus", "checked_at": "2026-10-11", "group": "interaction", "arena": "A09", "kind": "arena", "provider": "AMI Consortium", "release": "Manual annotations v1.6.2, 2017-04-10 (content unchanged since 2014-06-16); automatic annotations v1.5.1, 2017-04-10.", "units_and_scope": "English meeting recordings and synchronized annotations; approximately two-thirds scenario design meetings and one-third natural meetings.", "size": "About 100 hours; manual annotation ZIP 22 MB; automatic annotations 68 MB. Annotation types/coverage vary by session.", "coverage_and_dates": "Time-aligned word-level transcription, speaker turns and meeting sequences. Scenario projects comprise four meetings with four assigned roles.", "fields_and_links": "Meeting and speaker IDs, timed words, dialogue acts, topics and summaries; audio/video, slides and other meeting artifacts where available.", "access": "public download route; core CC BY 4.0 materials", "route": "Official corpus download selector and annotation archives.", "authentication": "Current annotation page exposes direct downloads; legacy overview describes registration. Current binary download was not tested.", "inspected": "Official overview, licensing notice and exact annotation-version entries inspected.", "reuse_status": "explicit CC BY 4.0 with component scope", "reuse_terms": "Main site licenses signals, transcriptions and some annotations CC BY 4.0. Check add-on annotation terms separately; do not extend that statement to every derived layer.", "reuse_group": "explicit", "limitations": [ "There is no AI treatment; scenario meetings are not natural firm decisions.", "Public talk does not reveal unspoken intentions, off-meeting work, actual approval powers or latent learning.", "Existing dialogue-act labels can assist retrieval but do not validate a new organizational construct." ], "suggested_task": "Trace proposal, evaluation, disagreement and recorded decision episodes. Link coding to timestamps, preserve contradictory turns and distinguish assigned roles from exercised authority." }, { "id": "D13", "name": "WikiConv", "checked_at": "2026-10-11", "group": "interaction", "arena": "A10", "kind": "arena", "provider": "Conversation AI", "release": "Creator corpus reconstructed from Wikipedia dumps dated 2018-07-01; creator repository read-only/archived 2026-09-04. ConvoKit supplies year/language subsets with their own representations.", "units_and_scope": "Wikipedia talk-page conversational actions and reconstructed discussion relationships across English, German, Russian, Chinese and Greek.", "size": "Creator English corpus reports 241,288,668 actions and 90,930,244 conversations under its technical definition. ConvoKit's documented English-2003 example has 140,265 utterances and 91,787 conversations; these are not interchangeable counts.", "coverage_and_dates": "Timestamps, revision/action histories and reply relationships support temporal reconstruction. Final utterance text can include later modifications.", "fields_and_links": "Action/utterance, speaker, conversation, page/revision and reply/parent identifiers; text, time, modification/restoration metadata. ConvoKit's representation must be distinguished from raw actions.", "access": "public archives; mixed CC0 and CC BY-SA 3.0 content", "route": "Creator Figshare/Google Cloud links or documented ConvoKit year subsets.", "authentication": "No research-approval requirement stated; no archive downloaded.", "inspected": "Creator README/license and ConvoKit documentation inspected.", "reuse_status": "component-specific explicit licenses", "reuse_terms": "Creator labels the dataset CC0 while individual comments remain CC BY-SA 3.0. Preserve comment attribution/share-alike obligations; ConvoKit software licensing does not replace content terms.", "reuse_group": "explicit", "limitations": [ "A technical conversation can contain one user; it is not automatically a collaborative episode.", "Volunteer discussion differs from employment-based organizing.", "Archival maintenance status, later-edited text and platform norms limit transfer to contemporary human–AI teams." ], "suggested_task": "Compare conclusions drawn from final text with those drawn from edits, reverts and reply timing; document how a changed state alters an episode interpretation." }, { "id": "D14", "name": "Noy–Zhang AI productivity experiment replication", "checked_at": "2026-10-11", "group": "interaction", "arena": "A11", "kind": "arena", "provider": "Open Science Framework", "release": "OSF project DOI 10.17605/OSF.IO/XD7QW; project metadata last modified 2023-07-21T14:22:07.394749. Public file inventories inspected; file-specific versions matter.", "units_and_scope": "Experiment with 453 college-educated professionals performing writing tasks, linked survey/essay data, external grades and follow-up surveys.", "size": "Final study N=453. OSF inspected inventories contain 12 raw-data files and 11 code files; those are package counts, not the analysis sample for every result.", "coverage_and_dates": "Two tasks, minute-level essay snapshots, self-reported task stages, and two-week/two-month follow-ups described in package README.", "fields_and_links": "Hashed participant IDs, treatment/order/incentive variables, task times, essay text snapshots and word-change counts; grader writing/content/originality judgments and justifications. Merge eligibility and analysis sample must follow the original specification.", "access": "public archive and file metadata; no declared license", "route": "OSF native API/file links; UI rendering was unreliable but public inventories and README were readable.", "authentication": "Project public=true, read permission confirmed; no login used.", "inspected": "README PDF inspected in memory and raw/code directory pagination exhausted; actual data and analysis code not acquired or executed.", "reuse_status": "unknown", "reuse_terms": "OSF native metadata returned node_license=null; no dataset license established by the inspected README. Public file links do not establish a redistribution grant.", "reuse_group": "unresolved", "limitations": [ "Stata Master.do and a Python analysis component require a later environment/dependency check; fixed paths and the revised grade-cleaning file need inspection.", "Minute-level text evolution does not reveal which AI outputs were seen, rejected or selected, nor stopping reasons.", "Follow-up AI use is not delayed unaided evaluative capability. Do not import older working-paper N=444 into the final study." ], "suggested_task": "Reproduce Figure 1 task-time and grade contrasts from the correct final-paper package. Trace discrepancies to sample, grade cleaning, text truncation, task order, software and specification." }, { "id": "D15", "name": "Hybrid working from home experiment replication", "checked_at": "2026-10-11", "group": "structured", "arena": "", "kind": "structured", "provider": "Harvard Dataverse", "release": "Harvard Dataverse replication DOI 10.7910/DVN/6X4ZZL, version 1.1, updated 2024-05-20 according to DataCite registration.", "units_and_scope": "Trip.com hybrid-work randomized experiment and anonymized outcome files described by the Nature article.", "size": "Study N=1,612 workers. Public package file counts, sizes and exact field inventory not verified.", "coverage_and_dates": "Six-month experiment August 2021–January 2022; article reports performance follow-up spanning July 2021–June 2023.", "fields_and_links": "Article describes assignment, attrition, performance reviews, promotion, surveys and a code-output subset. Which fields and joins are exposed in the public package remains to verify.", "access": "explicit CC BY 4.0 metadata; file access unverified", "route": "Harvard Dataverse DOI deposit.", "authentication": "Unknown for files. Dataverse UI/native routes failed or returned 403 in this inspection; DOI metadata was readable.", "inspected": "Nature data-availability statement and DataCite release/rights metadata inspected. No public file inventory or download verified.", "reuse_status": "CC BY 4.0 declared for deposit", "reuse_terms": "DataCite rightsList explicitly identifies CC BY 4.0. File-specific conditions and availability still need to be inspected on the deposit.", "reuse_group": "explicit", "limitations": [ "The article says data are split/anonymized to reduce reidentification; do not reconstruct an unrestricted person-level panel across deliberately separated files.", "This is not an AI experiment and does not reveal aspirations or internal coordination decisions.", "DOI licensing metadata confirms a declared license, not successful acquisition or runnable replication." ], "suggested_task": "Reconstruct the reported retention contrast and performance analysis, identifying which anonymized files supply each result." }, { "id": "D16", "name": "QASPER", "checked_at": "2026-10-11", "group": "products", "arena": "A06", "kind": "arena", "provider": "Allen Institute for AI", "release": "Creator HF revision fdc9d8214fbab5dd782958601db4d678e6934a54, modified 2022-10-07. A repository revision does not pin every external file used by its loader.", "units_and_scope": "Questions and evidence-linked answers about NLP research papers.", "size": "5,049 questions over 1,585 papers.", "coverage_and_dates": "Fixed benchmark content; no organizational event panel.", "fields_and_links": "Paper IDs, title/abstract, structured full-text paragraphs, questions, answer types, evidence passages and figure/table captions or references. Includes unanswerable and yes/no questions.", "access": "public benchmark; explicit CC BY 4.0", "route": "AllenAI creator HF card/files and paper-linked release.", "authentication": "gated=false in native metadata; no loader executed or data downloaded.", "inspected": "Creator card, native version metadata and paper inspected.", "reuse_status": "CC BY 4.0 dataset declaration", "reuse_terms": "Creator card declares CC BY 4.0. Separate source PDFs/additional article content require their own terms; Apache-2.0 baseline software license is distinct.", "reuse_group": "explicit", "limitations": [ "This is a supporting benchmark, not a substantive organizational dataset or validation of organizational constructs.", "Well-known public benchmark exposure can confound current-model comparisons.", "Evidence-match success does not establish correct downstream inference or genuine scientific discovery." ], "suggested_task": "Compare ordinary answering with answer-plus-evidence checking and explicit abstention." }, { "id": "D17", "name": "CFPB Consumer Complaint Database — current structured release", "checked_at": "2026-10-11", "group": "structured", "arena": "", "kind": "narrative-excluded", "provider": "Consumer Financial Protection Bureau", "release": "Live daily database; Release 24 (September 2026) removed complaint narratives and visualizations. Release 23 (July 2026) capped filtered CSV exports at 100,000 and retired JSON exports.", "units_and_scope": "Published consumer complaints and structured company-response categories.", "size": "Live population; no current exact row count verified. Download-export limits are not total population counts.", "coverage_and_dates": "Received/sent dates; historical coverage varies by product. Taxonomies and released fields change.", "fields_and_links": "Complaint ID, product/subproduct, issue/subissue, company, received/sent dates, submitted-via channel, state/ZIP and response/timeliness fields. Current public text narratives are absent.", "access": "public structured data; explicit free-use statement", "route": "Official public CSV exports and documented API, subject to current release changes.", "authentication": "Public routes have no account requirement stated; no complaint records acquired.", "inspected": "Current main page, API fields and September-2026 release notes inspected.", "reuse_status": "official free-use statement", "reuse_terms": "Main page states published data are freely available for use, analysis and building on; no standardized CC license was identified. Follow current source and sharing guidance.", "reuse_group": "explicit", "limitations": [ "Current acquisition cannot supply the former narrative-text task. Unverified mirrors are not a substitute for the current release.", "Published complaints are selected reports with no market-exposure denominator; rates are not population failure probabilities.", "Company public responses are preset categories, not free-form conversations; consent/disputed fields also changed in June 2026." ], "suggested_task": "Validate date/grouping/timeliness aggregation and show how a pipeline expecting narrative text detects the removed field rather than silently changing its task." }, { "id": "D18", "name": "UCI Online Retail", "checked_at": "2026-10-11", "group": "structured", "arena": "", "kind": "structured", "provider": "UCI Machine Learning Repository", "release": "UCI deposit donated 2015-11-05; DOI 10.24432/C5BW33; named file Online Retail.xlsx. No binary checksum acquired.", "units_and_scope": "Transaction line items from one UK online retailer, including many wholesale customers.", "size": "541,909 rows; workbook listed as 22.6 MB.", "coverage_and_dates": "2010-12-01 through 2011-12-09; timestamped invoices.", "fields_and_links": "InvoiceNo, StockCode, Description, Quantity, InvoiceDate, UnitPrice, CustomerID, Country. Invoice numbers beginning with C indicate cancellations; product and customer identifiers permit grouping.", "access": "public-download; explicit CC BY 4.0", "route": "Official UCI workbook download.", "authentication": "No login required by the displayed route; workbook not acquired.", "inspected": "Official creator deposit page inspected, including variables, file and license.", "reuse_status": "explicit CC BY 4.0", "reuse_terms": "UCI explicitly permits sharing/adaptation with attribution under CC BY 4.0.", "reuse_group": "explicit", "limitations": [ "One firm's transactions do not support a multi-firm theory test or customer-exposure denominator.", "Repository metadata's missingness claim is not a substitute for profiling the workbook; no data-level completeness was verified.", "A deterministic business-data control isolates computation failures but does not establish the validity of text-based organizational measures." ], "suggested_task": "Calculate a predeclared sales/return summary and identify how cancellation treatment, missing customer IDs and line-versus-invoice units change the conclusion." }, { "id": "D19", "name": "CoAuthor human–AI collaborative writing", "checked_at": "2026-10-11", "group": "interaction", "arena": "A11", "kind": "arena", "provider": "CoAuthor Project", "release": "CHI 2022 dataset; primary paper arXiv:2201.06796v2 (2022-01-25). Creator links to Drive session files and Google Sheets metadata/surveys; a versioned binary manifest was not exposed/inspected.", "units_and_scope": "English writing sessions by 63 crowd writers interacting with four instances of GPT-3.", "size": "1,445 sessions: 830 creative stories and 615 argumentative essays. File bytes not verified.", "coverage_and_dates": "Timestamped interaction/keystroke sequences within sessions. Collection calendar range and cross-session identifiers require release-file verification.", "fields_and_links": "Session events capture writing, cursor movement, requested suggestions, acceptance/dismissal and subsequent edits. Session metadata and surveys are linked from the creator site; exact raw key names/joins remain to inspect.", "access": "public creator download links; dataset reuse unresolved", "route": "Creator Drive files, metadata/survey sheet and session replay interface.", "authentication": "No account requirement stated on the landing page; binary/Sheet access not tested.", "inspected": "Creator overview and primary paper inspected; separate interface repository verified.", "reuse_status": "unknown for dataset", "reuse_terms": "No explicit data-specific reuse grant established. The interface is MIT-licensed and the arXiv paper has CC BY 4.0; neither automatically licenses session content, survey data or source prompts. Third-party license labels were not adopted.", "reuse_group": "unresolved", "limitations": [ "Suggestion display does not prove attention, comprehension or a change in private criteria; session termination does not explain stopping motives.", "Crowd writing with 2022-era systems does not establish contemporary workplace coordination, durable skill or multi-agent authority.", "Do not silently equate the interface software release with the original dataset or use derivative-repository licenses as source-data permission." ], "suggested_task": "Reconstruct candidate→displayed suggestion→accept/dismiss→edit paths and compare them with conclusions from final text alone. Distinguish a recorded interface action from its interpretation." }, { "id": "D20", "name": "Doshi–Hauser creativity experiment replication", "checked_at": "2026-10-11", "group": "interaction", "arena": "A11", "kind": "arena", "provider": "Dryad", "release": "Dryad release dated 2024-06-14; DataCite version 6. Related Zenodo record 11654403 reports the same Dryad DOI and CC0. ZIP md5 b329200a3bab35b5326fdd6c7191eaa8.", "units_and_scope": "Randomized short-story writing experiment with separate evaluators and post-disclosure ownership judgments.", "size": "293 stories/writers; 600 evaluators and 3,519 retained evaluations. Archive lists a 2,472,080-byte scripts/data ZIP and 3,657-byte README.", "coverage_and_dates": "One experimental writing/evaluation sequence; release date is not collection date. No delayed unaided follow-up established.", "fields_and_links": "Conditions: human-only, access to one AI idea, access to up to five ideas; story content, writer assessments, evaluator novelty/usefulness judgments and ownership responses. Exact variable names and story/evaluator merge keys require file inspection.", "access": "public archive; explicit CC0", "route": "Dryad canonical deposit and related Zenodo distribution; README describes raw CSV and processed Stata data.", "authentication": "Public listing/download links, Zenodo access_right=open. No files acquired.", "inspected": "Deposit README rendered in-page; native DataCite and Zenodo metadata checked.", "reuse_status": "CC0 1.0", "reuse_terms": "DataCite rightsList and Zenodo metadata explicitly identify CC0 for this deposit. Separate dependencies and software retain their own terms.", "reuse_group": "explicit", "limitations": [ "The full rebuild requires Python, external language resources and an OpenAI API key; live calls may not reproduce historical embeddings/results. Processed-data analysis avoids that rebuild but still needs Stata/dependencies.", "README mentions GenAI_creativity_data_and_scripts.zip while the listed file is GenAI_creativity_scripts.zip; inspect archive inventory before claiming runnable reproduction.", "Creativity ratings are domain-specific judgments. Better rated stories, collective diversity and durable unaided capability are different outcomes." ], "suggested_task": "Follow the documented processed-data analysis path first, then separately diagnose which conclusions depend on rebuilding similarity/preprocessing inputs. Reconcile assignment, actual idea use and exclusions." }, { "id": "D21", "name": "BPI Challenge 2017 loan-application event log", "checked_at": "2026-10-11", "group": "structured", "arena": "", "kind": "structured", "provider": "4TU.ResearchData", "release": "4TU dataset version 1, issued 2017-02-06; archive item 12696884. Canonical DOI metadata checked; Figshare-hosted record rendered.", "units_and_scope": "Applications, multiple offers per application, and operational workflow events at one Dutch financial institution.", "size": "Archive lists 28.29 MB compressed. A published challenge analysis reports 31,509 applications, 42,995 offers and 149 originators; originators include employees or systems, not necessarily distinct workers.", "coverage_and_dates": "Applications filed during 2016; subsequent events through 2017-02-01 per archive. Timestamp/lifecycle interpretation and end-of-window censoring need checking.", "fields_and_links": "Application and offer IDs; application type, requested amount and loan goal; offer amounts/terms, selected/accepted indicators, credit score and creating employee. Application/offer/workflow event classes permit case trajectories; verify actual XES attribute names and lifecycle fields.", "access": "public archive; custom legacy terms not fully resolved", "route": "4TU canonical DOI and linked Figshare record, compressed XML/XES distribution.", "authentication": "Public download displayed; no archive acquisition attempted.", "inspected": "Native DOI metadata, deposit description and primary challenge analysis inspected. Legacy DOI/terms resolver and organizer page failed in the browser.", "reuse_status": "4TU General Terms of Use; operative text unresolved", "reuse_terms": "Deposit and DataCite label the original release with 4TU General Terms of Use. Do not substitute CC BY 4.0 from a later derived event-graph release or Apache/MIT from analysis code. Inspect operative terms before redistribution.", "reuse_group": "unresolved", "limitations": [ "Elapsed time between events is not hands-on labor; suspend/resume and parallel work can defeat simple subtraction.", "Resource IDs do not reveal actual authority, attention, skill or contractual roles; this has no AI treatment.", "The 2012 log is a related earlier system, not an interchangeable release; a system change is not automatically exogenous." ], "suggested_task": "Construct NEW case-level handoff/rework observations, maintaining application–offer relations. Compare a naive timestamp summary with lifecycle-aware reconstruction." }, { "id": "D22", "name": "O*NET 31.0 occupation, task and work-activity database", "checked_at": "2026-10-11", "group": "institutional", "arena": "A03", "kind": "arena", "provider": "National Center for O*NET Development", "release": "Production version 31.0, August 2026; prior versions remain in the release archive. Pin 31.0 rather than a mutable current-data endpoint.", "units_and_scope": "Occupation-specific task statements and aggregate ratings, linked to skills, work activities and work context.", "size": "31.0 lists 18,838 task statements and 165,780 task-rating rows; ratings cover 911 occupations. Occupation-data table has 1,016 rows, which is a different coverage denominator.", "coverage_and_dates": "Release versions and field-update dates; a quarterly release is not a repeated survey of every occupation. Avoid interpreting asynchronous updates as real work change.", "fields_and_links": "O*NET-SOC code + Task ID link statements and ratings; task type, source and update date; ratings include scale/category, value, N, uncertainty and suppression flags. Tasks-to-DWA maps provide a further work-activity join.", "access": "public static downloads; CC BY 4.0", "route": "Official Excel/CSV/JSON, SQL and RDF files with release-specific dictionaries.", "authentication": "Static downloads are public; developer registration is optional. Web Services have a separate sign-up route.", "inspected": "31.0 release/archive, task dictionaries and license inspected; files not downloaded.", "reuse_status": "CC BY 4.0 for specified downloadable database files", "reuse_terms": "Attribute the database version and USDOL/ETA, link the license and identify modifications. Trademark rules and separate career-exploration-tool licenses still apply; database terms do not license every O*NET website/tool.", "reuse_group": "explicit", "limitations": [ "Occupation aggregates are not observed employees, establishments or workflows.", "Task descriptions do not establish AI adoption, feasible automation, time saved, authority transfer or job displacement.", "A new questionnaire requires respondent testing; reusing O*NET wording does not validate the new instrument." ], "suggested_task": "Develop a source-linked coding scheme for information needs, review requirements and possible human/AI allocation. Keep inferred allocation labels separate from official task attributes." }, { "id": "D23", "name": "GLEIF Golden Copy legal entities and relationships", "checked_at": "2026-10-11", "group": "structured", "arena": "", "kind": "structured", "provider": "GLEIF", "release": "Live Golden Copy files issued three times daily; schema documentation lists LEI-CDF 3.1, Relationship Record CDF 2.1 and Reporting Exceptions 2.1. No timestamped file release acquired.", "units_and_scope": "LEI-identified legal entities, direct/ultimate accounting-consolidating parent relationships and explicit reporting exceptions.", "size": "Current exact record count/bytes not verified. Choose a dated file manifest or bounded API result in later acquisition.", "coverage_and_dates": "Registration/update information and current relationship records; delta windows of 8 hours, 24 hours, 7 days and 31 days. A database change is not necessarily the date of an ownership event.", "fields_and_links": "LEI links entity reference data and parent/child relationships; legal names, addresses, mapped identifiers where available and exceptions. Missing parent LEI, no parent and non-reporting reasons must remain distinct.", "access": "public, no registration; CC0", "route": "Official Golden Copy/delta files and API; API offers field/name searches and relationship retrieval.", "authentication": "GLEIF documents free access without registration; no data acquired.", "inspected": "Official release/schema, API/access and terms pages inspected.", "reuse_status": "CC0 1.0", "reuse_terms": "GLEIF's LEI Data Terms of Use explicitly provide data through the Access Service under CC0. Preserve source/version provenance for research even where attribution is not a license condition.", "reuse_group": "explicit", "limitations": [ "Accounting-consolidating parenthood is not all beneficial ownership, voting control or everyday decision authority.", "Entities with LEIs are a selected population, not a census of firms.", "Current Golden Copy relationships do not supply a full historical ownership panel; examples in schema documentation are not observed validation cases. Do not assume an automatic CIK–LEI bridge." ], "suggested_task": "Construct a new adjudicated entity-resolution/parent-reporting table, optionally matched to SEC issuers where a defensible mapping exists. A later change dataset would require historical versions and event-source corroboration." }, { "id": "D24", "name": "SocioPatterns contacts in a workplace", "checked_at": "2026-10-11", "group": "structured", "arena": "", "kind": "structured", "provider": "SocioPatterns", "release": "Creator release posted 2016-06-23; two named files on the dataset page. File hashes were not obtained.", "units_and_scope": "Anonymous person-pair proximity/contact intervals and workplace departments in one French office.", "size": "Creator lists 153 KB contact file and 0.8 KB department file. Exact participant/event counts were not independently verified in this round.", "coverage_and_dates": "2013-06-24 to 2013-07-03; 20-second intervals measured relative to midnight on June 24.", "fields_and_links": "Contact table t,i,j; department table i,Di. Join anonymous person ID i (or j) to department; interval represented as [t−20 seconds,t].", "access": "public direct files; CC0", "route": "Creator tab-separated contact and department files.", "authentication": "Direct downloads shown with no login; no records downloaded.", "inspected": "Creator schema/period/size/terms page and linked CC0 deed inspected.", "reuse_status": "CC0 1.0", "reuse_terms": "Creator explicitly links the Creative Commons Public Domain Dedication; requests citation to the associated Network Science paper.", "reuse_group": "explicit", "limitations": [ "Co-presence/contact does not reveal speech content, cooperation, information exchange or decision rights.", "One short office observation cannot establish enduring organizational structure or an AI effect.", "Anonymity does not justify reidentification; department/temporal coverage may be selective. No clinical or vaccination recommendation is proposed." ], "suggested_task": "Build an interdepartmental contact-duration network and compare justified temporal aggregation rules. Keep the output labeled recorded contact, not collaboration." }, { "id": "D25", "name": "WildChat writing-interaction annotations (Mysore et al.)", "checked_at": "2026-10-11", "group": "interaction", "arena": "A11", "kind": "arena", "provider": "Microsoft", "release": "Creator repository ID 992823885; pinned commit 65b858edb2be77bca417cafdb06a2c4cb4c113b5, 2025-09-19. Derived from WildChat-1M; do not count the parent as another new family here.", "units_and_scope": "Observed user–LLM conversation sessions with machine-produced task, follow-up and writing-intent labels.", "size": "README lists 159,134 selected sessions, 68,782 writing sessions and 32,697 final analysis sessions after annotation-alignment filtering. These are nested stages, not additive samples.", "coverage_and_dates": "Turn order and conversation timestamp; README specifies timestamp as last turn in UTC, not an event time for every user action.", "fields_and_links": "convid, ordered turns(author,utterance), model and timestamp; coarse_tasks, utterance_types and writing_intents. Repository also includes IP-derived fields; they are unnecessary for the proposed task and should be omitted.", "access": "public creator repository; ODC-BY dataset declaration", "route": "Three documented JSONL-related files in data/ at the pinned commit.", "authentication": "Public GitHub metadata and README readable without login; data files not acquired.", "inspected": "Native repo/commit metadata and exact-commit README inspected; primary paper checked.", "reuse_status": "ODC-BY 1.0 declared for data", "reuse_terms": "README explicitly licenses data ODC-BY. This database-level license should not be treated as clearance for every user-authored utterance; retain upstream provenance and inspect sharing/privacy scope before distributing raw conversations.", "reuse_group": "explicit", "limitations": [ "GPT-4o labels are model outputs, not human ground truth. The README's three-utterance selection wording needs reconciliation with the paper/actual records before sampling.", "Only public WildChat material is released; do not assume access to every source used in the paper.", "README still marks analysis/data-generation code as TODO. Conversation endings do not establish stopping reasons; logs do not show copied text, external editing, attention or actual workplace roles." ], "suggested_task": "Recode user requests and revisions using a human-checked event dictionary; compare inferred collaboration patterns before and after correcting labels and selection." } ], "provider_checks": [ { "title": "FederalRegister.gov REST API", "url": "https://www.federalregister.gov/reader-aids/developer-resources/rest-api", "checked_claim": "CSV/JSON; no API key; first 2,000-result limit and date partitioning; informational web edition notice." }, { "title": "GovInfo Federal Register help", "url": "https://www.govinfo.gov/help/fr", "checked_claim": "Office of the Federal Register/NARA and GPO provider roles; official publication description." }, { "title": "SEC EDGAR APIs", "url": "https://www.sec.gov/search-filings/edgar-application-programming-interfaces", "checked_claim": "Filing histories by issuer and structured financial information; links to filings." }, { "title": "USPTO PatentsView", "url": "https://www.uspto.gov/ip-policy/economic-research/patentsview", "checked_claim": "PatentsView now hosted in the USPTO Open Data Portal." }, { "title": "CFPB complaint API release notes", "url": "https://cfpb.github.io/api/ccdb/release-notes.html", "checked_claim": "September 2026 release removed narratives; current database not recommended as a complaint-narrative source." }, { "title": "CoAuthor creator page", "url": "https://coauthor.stanford.edu/", "checked_claim": "1,445 human–AI writing sessions; suggestions, acceptance/dismissal and editing traces." }, { "title": "AMI provider corpus page", "url": "https://groups.inf.ed.ac.uk/ami/corpus/", "checked_claim": "Meeting speech/transcripts and annotations; scenario-based and naturally occurring meetings differ." }, { "title": "Open Food Facts API", "url": "https://openfoodfacts.github.io/openfoodfacts-server/api/", "checked_claim": "Barcode identifiers, contributed fields and label images; components carry distinct reuse terms." }, { "title": "Amazon Reviews 2023 creator page", "url": "https://amazon-reviews-2023.github.io/", "checked_claim": "Review text, item metadata and user–item links in the 2023 release." }, { "title": "O*NET database", "url": "https://www.onetcenter.org/database.html", "checked_claim": "Version 31.0; occupation/task data files and dictionaries." } ] }