Rich textual-data sources review

Research foundation

A representative survey of textual sources and complementary structured data.

Choose by what a source records

A searchable catalogue and 25 dated profiles make source coverage, linking fields and reuse limits visible before acquisition. The working renderer turns the included records into readable profiles, offline.

One missing field changes the choice

The consumer-complaint database looked promising for complaint–response research. Its profile records that the provider’s September 2026 release removed complaint narratives. Structured fields remain, but they cannot support the proposed narrative task.

Google Local is the next candidate: it documents review text and dated business responses. Its reuse terms remain unresolved. The completed result is a better-informed selection decision; acquisition and the empirical task remain proposed.

Compare the profiles: the complaint database · Google Local.

Read the AI prompt

# Explore the textual-data sources prototype

Read the public page at https://textdata.info/foundation/textual-data-sources/ and the component folder at https://github.com/joyzhzh/textdata/tree/main/appendix/04-data-sources.

Start with `README.md`, `docs/comparison.md`, `example/output/D17.md` and `data/profiles/D02.md`. Explain the concrete selection example: CFPB’s September 2026 release removed complaint narratives, while Google Local documents review text and dated business responses but has unresolved reuse terms. Explain what that changes for a complaint–response research question.

Help me explore the 25 profiles using the page’s search and reuse filters, or compare relevant entries in `data/catalogue.json`. Use `data/evidence.jsonl` and `data/provider_checks.json` for source locators and dates. If I want to run the prototype, use the documented Python commands in `README.md` and explain the output from `tool/render_profile.py`.

Distinguish the completed catalogue, comparative review and working renderer from proposed acquisition, empirical tasks and study validation. Do not imply that source datasets are included or that public access establishes reuse permission. If you cannot access a file or run a command, say so.

My research question or source-selection problem: [describe it here].

What has been built

A catalogue, dated profiles, comparative review and profile-rendering tool describe coverage and observability. A change in released fields altered a proposed source selection.

What comes next

Update profiles, assess access and suitability for new questions, and guide further acquisition.

What works now

18 in twelve arenas of rich textual data
6 structured complements
1 no longer a narrative source

Explore the profiles below, download the catalogue, or run the profile renderer from the repository. The comparative review explains which sources suit different demonstrations.

Profiles are dated 11 October 2026; ten provider pages were checked on 29 October. These are documented source options, not datasets included here.

Explore the catalogue

Choose the kind of record a question needs or the reuse terms a study can accept, and open a family to read its profile.

Search the catalogue

Showing 25 of 25 families

Download the catalogue: CSV JSON

Institutional records and organizational claims

Regulation and public administration

Framing and stated obligations; text is not implementation or behavior.

Sources: Federal Register

Federal Register documents and metadataOffice of the Federal Register and Government Publishing Office · public access; government-content reuse with exclusions

Provider: Office of the Federal Register and Government Publishing Office
Coverage and dates: Publication, effective, comment and action dates are different. Date presence and legal status need document-level checking.
Reuse: government public-domain component; inspect exceptions
Profile: D10 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

The collection itself is on the Large-scale data acquisition page.

Corporate disclosure

Issuer claims and disclosed events; disclosure and decision dates differ.

Sources: SEC EDGAR

SEC EDGAR filings and company factsU.S. Securities and Exchange Commission · public access with official dissemination permission

Provider: U.S. Securities and Exchange Commission
Coverage and dates: Filing dates, reporting periods and event dates are distinct. XBRL required from 2009; amendments and restatements matter for as-of analyses.
Reuse: official copying/distribution permission, with exclusions
Profile: D09 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

Work and occupations

Task descriptions and instrument design; no direct evidence of automation or authority.

Sources: O*NET; job postings

O*NET 31.0 occupation, task and work-activity databaseNational Center for O*NET Development · public static downloads; CC BY 4.0

Provider: National Center for O*NET Development
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.
Reuse: CC BY 4.0 for specified downloadable database files
Profile: D22 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

Products, innovations and evaluations

Product launches and descriptions

Described functionality; posts, containers and companies are different units.

Sources: Product Hunt

No family of the catalogue sits in this arena.

Patents

Technological comparison sets; patents do not cover all innovation.

Sources: USPTO PatentsView

USPTO PatentsView disambiguated patent dataU.S. Patent and Trademark Office · registration-controlled public source; current release terms unresolved

Provider: U.S. Patent and Trademark Office
Coverage and dates: Application, publication and grant dates have different meanings. Historical reference sets must use information actually public before each target cutoff.
Reuse: manifestation-dependent; current terms unverified
Profile: D08 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

Scientific text and knowledge links

Abstract/reference analysis and evidence-grounding fixtures; incomplete fields and full-text rights.

Sources: OpenAlex; QASPER

OpenAlex scholarly graphOpenAlex · CC0 metadata; API account/key for regular use

Provider: OpenAlex
Coverage and dates: Publication dates and snapshot dates differ; entity resolution, affiliations and classifications can change.
Reuse: CC0 metadata
Profile: D11 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

QASPERAllen Institute for AI · public benchmark; explicit CC BY 4.0

Provider: Allen Institute for AI
Coverage and dates: Fixed benchmark content; no organizational event panel.
Reuse: CC BY 4.0 dataset declaration
Profile: D16 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

Product composition and artifacts

Attribute/recombination comparisons; snapshot novelty is not historical firstness.

Sources: Open Food Facts; Food.com; RecipePairs

Food.com Recipes and User InteractionsKaggle · public listing; download authentication unverified; reuse restricted/unclear

Provider: Kaggle
Coverage and dates: Recipe submitted dates and interaction dates allow a declared temporal reference set; exact date completeness remains untested.
Reuse: copyright notice; no broad reuse grant found
Profile: D03 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

RecipePairs / SHARE recipe-editing dataHugging Face · public-download; declared GPL-3.0 with upstream-rights caveat

Provider: Hugging Face
Coverage and dates: Pair-table dates were not verified; historical novelty requires a checked join to original recipe dates.
Reuse: declared GPL-3.0; upstream provenance unresolved
Profile: D04 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

Open Food FactsOpen Food Facts · public access; explicitly licensed database

Provider: Open Food Facts
Coverage and dates: Record creation/modification timestamps concern database observations, not product launch dates or invention.
Reuse: explicit, component-specific
Profile: D07 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

Reviews and responses

Evaluative categories and response sequences; selection and reuse limits remain.

Sources: Amazon Reviews 2023; Google Local; Goodreads

Amazon Reviews 2023McAuley Lab · public-download; reuse unclear

Provider: McAuley Lab
Coverage and dates: Review dates May 1996–September 2023. Product metadata is a snapshot, not a historical price or product-attribute panel.
Reuse: unknown
Profile: D01 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

Google Local Data 2021McAuley Lab · public-download; reuse unclear

Provider: McAuley Lab
Coverage and dates: Review time and response time are available. Business attributes are snapshot values; collection coverage is not an operating-history panel.
Reuse: unknown
Profile: D02 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

Goodreads / UCSD Book GraphMengting Wan · academic-use-only; current download unverified

Provider: Mengting Wan
Coverage and dates: Public interaction/review dates where present; edition publication and work identity require separate treatment.
Reuse: restricted research reuse
Profile: D05 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

Interaction and process records

Meetings and interview accounts

Situated discussion or mediated accounts; source types are not interchangeable.

Sources: AMI; publisher podcast transcripts

AMI Meeting CorpusAMI Consortium · public download route; core CC BY 4.0 materials

Provider: AMI Consortium
Coverage and dates: Time-aligned word-level transcription, speaker turns and meeting sequences. Scenario projects comprise four meetings with four assigned roles.
Reuse: explicit CC BY 4.0 with component scope
Profile: D12 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

Collaborative discussion and platforms

Contribution and ecosystem histories; current observations are not complete history.

Sources: WikiConv; Hugging Face

WikiConvConversation AI · public archives; mixed CC0 and CC BY-SA 3.0 content

Provider: Conversation AI · ConvoKit documentation · paper
Coverage and dates: Timestamps, revision/action histories and reply relationships support temporal reconstruction. Final utterance text can include later modifications.
Reuse: component-specific explicit licenses
Profile: D13 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

Human–AI writing and creative traces

Writing/generation sequences and comparisons; private criteria and stopping reasons remain unobserved.

Sources: CoAuthor; Noy–Zhang; Doshi–Hauser; WildChat annotations

Noy–Zhang AI productivity experiment replicationOpen Science Framework · public archive and file metadata; no declared license

Provider: Open Science Framework
Coverage and dates: Two tasks, minute-level essay snapshots, self-reported task stages, and two-week/two-month follow-ups described in package README.
Reuse: unknown
Profile: D14 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

CoAuthor human–AI collaborative writingCoAuthor Project · public creator download links; dataset reuse unresolved

Provider: CoAuthor Project
Coverage and dates: Timestamped interaction/keystroke sequences within sessions. Collection calendar range and cross-session identifiers require release-file verification.
Reuse: unknown for dataset
Profile: D19 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

Doshi–Hauser creativity experiment replicationDryad · public archive; explicit CC0

Provider: Dryad
Coverage and dates: One experimental writing/evaluation sequence; release date is not collection date. No delayed unaided follow-up established.
Reuse: CC0 1.0
Profile: D20 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

WildChat writing-interaction annotations (Mysore et al.)Microsoft · public creator repository; ODC-BY dataset declaration

Provider: Microsoft · paper
Coverage and dates: Turn order and conversation timestamp; README specifies timestamp as last turn in UTC, not an event time for every user action.
Reuse: ODC-BY 1.0 declared for data
Profile: D25 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

Text on four platforms

Descriptions, comments, captions and transcripts differ.

No family of the catalogue sits in this arena.

Complementary structured data

Structured complements

These six datasets supply tags, context, comparisons or identifiers rather than rich prose.

MovieLens 32MGroupLens · public-download; custom research terms

Provider: GroupLens
Coverage and dates: Ratings from 1995-01-09 through 2023-10-12; ratings and tags carry timestamps.
Reuse: custom noncommercial/research conditions
Profile: D06 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

Hybrid working from home experiment replicationHarvard Dataverse · explicit CC BY 4.0 metadata; file access unverified

Provider: Harvard Dataverse
Coverage and dates: Six-month experiment August 2021–January 2022; article reports performance follow-up spanning July 2021–June 2023.
Reuse: CC BY 4.0 declared for deposit
Profile: D15 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

UCI Online RetailUCI Machine Learning Repository · public-download; explicit CC BY 4.0

Provider: UCI Machine Learning Repository
Coverage and dates: 2010-12-01 through 2011-12-09; timestamped invoices.
Reuse: explicit CC BY 4.0
Profile: D18 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

BPI Challenge 2017 loan-application event log4TU.ResearchData · public archive; custom legacy terms not fully resolved

Provider: 4TU.ResearchData
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.
Reuse: 4TU General Terms of Use; operative text unresolved
Profile: D21 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

GLEIF Golden Copy legal entities and relationshipsGLEIF · public, no registration; CC0

Provider: GLEIF
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.
Reuse: CC0 1.0
Profile: D23 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

SocioPatterns contacts in a workplaceSocioPatterns · public direct files; CC0

Provider: SocioPatterns
Coverage and dates: 2013-06-24 to 2013-07-03; 20-second intervals measured relative to midnight on June 24.
Reuse: CC0 1.0
Profile: D24 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

No longer a narrative source

The consumer-complaint database is kept for its structured fields; its September 2026 release removed the complaint narratives.

CFPB Consumer Complaint Database — current structured releaseConsumer Financial Protection Bureau · public structured data; explicit free-use statement

Provider: Consumer Financial Protection Bureau
Coverage and dates: Received/sent dates; historical coverage varies by product. Taxonomies and released fields change.
Reuse: official free-use statement
Profile: D17 · 2026-10-11

The full profile is in the catalogue downloads (CSV, JSON).

Run or reuse it

The appendix includes the catalogue, 25 rendered profiles, comparative review, source evidence and a Python renderer. The README gives offline commands and a check that reproduces the shipped profiles.

Open the component folder

Limits

A representative documentation review: source datasets were not acquired, proposed tasks were not run, and study fit has not been validated.
Profiles record dated conditions. Releases, fields and access routes can change; check the provider before acquisition.
Public access does not establish reuse permission. Several sources have unresolved or restricted terms; the profiles retain those limits.

Credits

The profiles rest on the documentation of their providers: 4TU.ResearchData; Allen Institute for AI; AMI Consortium; CoAuthor Project; Consumer Financial Protection Bureau; Conversation AI; Dryad; GLEIF; GroupLens; Harvard Dataverse; Hugging Face; Kaggle; McAuley Lab; Mengting Wan; Microsoft; National Center for O*NET Development; Office of the Federal Register and Government Publishing Office; Open Food Facts; Open Science Framework; OpenAlex; SocioPatterns; U.S. Patent and Trademark Office; U.S. Securities and Exchange Commission; UCI Machine Learning Repository.

Licenses are named as the review recorded them from each provider’s terms.