HathiTrust Research Center Analytics

register entry hathitrust-research-center-aio-20261101 OBS-e56da33771d0 · file as first run sha256 8fcebde599d3d79f…

• T1-1 definition/status: 'The HathiTrust Research Center (HTRC) is winding down its public services as HathiTrust suspends HTRC funding at the end of 2026.' (two markers)
• T1-2 closure timeline: 'Closure Timeline: Announced in October 2024, funding for the HTRC concludes at the end of 2026, and the center is scheduled to officially close.' (two markers)
• T1-3 unavailable services: 'Unavailable Services: Public-facing HTRC services—including data capsules, analytics algorithms, and analytics gateway login—became unavailable as of September 30, 2026.' (two markers)
• T1-4 purpose and origin: 'Purpose & Origin: Founded in 2011 as a partnership between Indiana University, the University of Illinois Urbana-Champaign, and HathiTrust, the center enabled large-scale text and data mining of the HathiTrust Digital Library.' (two markers)
• T1-5 future direction: 'Future Direction: HathiTrust is reallocating resources toward emerging technologies and alternative modes of text/data access and machine learning support for its users.' (two markers)
• T1-6 closing offer: 'Would you like to know more about alternative ways to access the HathiTrust Digital Library collection or explore general library text-mining resources?' (no marker)

• B1 HathiTrust — HathiTrust Research Center – HathiTrust Digital Library unreachable: https://www.hathitrust.org/about/research-center/ HTTP 403 at 2026-11-01T15:13:30Z; card title and snippet only
• B2 HathiTrust — HTRC Analytics card and census own page; B2.html sha256 5e7d8be238f03fa7, HTTP 200 at 2026-11-01T04:07:50Z (the census's copy; this row's refetch at 15:13:55Z byte-identical)
• B3 HathiTrust — HathiTrust Research Center unreachable: https://www.hathitrust.org/about/research-center/ HTTP 403 at 2026-11-01T15:13:30Z; https://htrc.illinois.edu/ HTTP 200 at 2026-11-01T15:13:31Z (empty body, 0 bytes; again at 15:14Z); card title and snippet only
• B4 HathiTrust — Update on HathiTrust Research Center unreachable: https://www.hathitrust.org/press-post/update-on-hathitrust-research-center/ HTTP 403 at 2026-11-01T15:13:30Z; https://www.hathitrust.org/about/research-center/htrc-transition-faq HTTP 403 at 2026-11-01T15:13:31Z; https://www.hathitrust.org/press-post/plans-for-hathitrust-research-center/ HTTP 403 at 2026-11-01T15:13:31Z; card title and snippet only
• B5 HTRC Analytics — About (own page, fetched by this row) own page, fetched by this row; B5.html sha256 032e23f9e0697c13, HTTP 200 at 2026-11-01T15:13:55Z
• B6 HTRC Analytics — Derived datasets (own page, fetched by this row) own page, fetched by this row; B6.html sha256 e4bf2715eb11a778, HTTP 200 at 2026-11-01T15:14:13Z
• B7 HathiTrust Research Center — htrc.illinois.edu (census own page, empty) own page (census htrc--about): HTTP 200, empty body (0 bytes) at 04:07:50Z and on refetch; unread

E_A, the archive's entity: the same center, in the traininglayerliterature.org layer census record and the archive's deposits: an earlier receiving apparatus with its product on record, whole library volumes read by computers for scholars under a non-consumptive policy and what that reading made published as derived datasets that outlast the service, the inspectable form of the transformation, the volumes staying with the library; infrastructure to the census's question of how literature is addressed to a receiving apparatus and what that apparatus makes from it; in the census's area of the corpora with no model training on record, kept apart from training datasets and from writing addressed to a machine reader; #1543 defines the receiving apparatus and machine reception as an encounter that may emit an artifact; #186 holds that literary works in the corpora are read by machines incidentally; #700 and #400 read computational literary method as taking the text as data point and machine readership as method; #127 holds unattributed ingestion a provenance-stripped copy TLL-25 TLL-26 TLL-13 TLL-16 TLL-23 TLL-33 TLL-35 TLL-10 TLL-08 TLL-06 O1543-01 O1543-02 O1543-04 O186-02 O700-01 O400-01 O127-01; defining papers #1543, #186, #400, #700, #127

E_F, the field's entity (type C — a conventional reading holds it (public entity: a research center and its analytics service for computational analysis of library volumes)): the HathiTrust Research Center (HTRC) and HTRC Analytics, its portal: a research center established in 2011 as a partnership of Indiana University, the University of Illinois Urbana-Champaign and HathiTrust, co-located at the two universities, dedicated to enabling computational analysis of the HathiTrust Digital Library for educational and non-profit research under a non-consumptive research paradigm (released data minimal, unable to be easily reconstructed into the raw volume); it offered worksets, algorithms and data capsules and released derived datasets (Extracted Features for 17.1 million volumes under CC BY 4.0, BookNLP for fiction, word frequencies, geographic locations); HathiTrust announced in October 2024 the end of funding and data access at the end of 2026; its services ended on 30 September 2026, Bookworm, the derived datasets and a visualization remaining without an account; its team plans its own research after 2026

ρ = extends: The archive reads the center as its cards and own pages present it and adds, on the same object, the reading of its derived datasets as the recorded product of a machine reading of library volumes (TLL-25, TLL-26), the receiving apparatus that reading names (#1543: O1543-01, O1543-02, O1543-04), the question of model training, which the census leaves open and keeps apart (TLL-13, TLL-23, TLL-33), the place of literary volumes read by machines incidentally (#186: O186-02), the stance of computational method toward the text (#700, #400), and the question of a work's name in the derived data (TLL-14; #127). No archive claim contradicts a field claim. TLL-25 TLL-26 TLL-13 TLL-23 TLL-33 TLL-14 O1543-01 O1543-02 O1543-04 O186-02 O700-01 O400-01 O127-01. Guard: ρ = extends claims an addition and no more: the archive's claims do not make the field's sources wrong at their dates; the census's open question of model training stays a question (open block), and #127's rule is the archive's for its own works.

The case: equal terms, ontology included (v0.9). The archive reads the field's entity as the entity. The census record describes the center from its own home page and keeps the resource's terms (operator, kind, unit, record object, method, lifecycle) apart from its own reading (tll_reading), which reads the same center as an earlier receiving apparatus with its product on record and its derived datasets as the inspectable form of the transformation. Its anchors define the receiving apparatus (#1543) and the layer where texts are crawled, indexed, trained on, retrieved and recomposed (#186); #700 and #400 read computational literary method; #127 holds that unattributed ingestion produces a provenance-stripped copy. None of these reads the center as made by reception or contests that it is the 2011 partnership's center for non-consumptive computational analysis of the HathiTrust Digital Library: they add senses (the derived datasets as the recorded product of a machine reading of literature; the question of model training, which the census leaves open; the stance of computational method toward the text) on the field's object. ρ = extends; the equal-terms case.

#TLL census.json rows[slug == 'htrc'].tll_reading: The archive reads HTRC as an earlier receiving apparatus with its product on record: whole library volumes read by computers for scholars under a non-consumptive policy, and what that reading made (extracted features, BookNLP data, word frequencies, geographic locations) published as derived datasets that outlast the service.

#TLL census.json rows[slug == 'htrc'].tll_reading: The derived datasets are the inspectable form of the transformation; the volumes themselves stay with the library.

#TLL census.json rows[slug == 'htrc'].record.in_layer: the saved sources say nothing of model training

#1543 §8 The genre claim, retyped (line 208): Under TLL*, scripture is the genus's longest-running species, and the human institutional layers were the first receiving apparatus.

#1543 §10 Formal consolidation (lines 258–259): the encounter changes the machine state and may emit an artifact

#186 Who is in the field? ¶4 (line 54): They are read by machines incidentally.

equal-terms case: every field claim composed as received, beside the archive claims it stands with; no receiving text. Reading the received column recovers the field (§5.9).

v0.9 §4.10 (unchanged in v0.10): KO_B sentences surviving in the KO at similarity ≥ 0.95, with positions (computed as apply_v09.py does: difflib.SequenceMatcher ratio, the best KO sentence for each KO_B sentence): v0.10 0 of 10 (positions [None, None, None, None, None, None, None, None, None, None]; composed first under v0.10, so there is no v0.7 object to measure against). Opening-run check (§5.10): no breach.

The HathiTrust Research Center (HTRC) is a research center dedicated to enabling computational analysis of the HathiTrust Digital Library, supporting large-scale text analysis for educational and non-profit research. It was established in 2011 as a partnership between Indiana University, the University of Illinois Urbana-Champaign and HathiTrust, co-located at the two universities, with John Walsh as director and J. Stephen Downie as co-director. It developed tools, datasets and infrastructure for text and data mining of the HathiTrust corpus, organized as worksets, algorithms and data capsules. It operates under a non-consumptive research paradigm within the fair-use rights courts have recognized for text analysis, and its non-consumptive use policy releases only minimal data that cannot easily be reconstructed into the raw volume. HathiTrust, a partnership holding millions of digitized titles, serves individual access and reading, and the center the aggregate question of what can be learned from so many books. Its derived datasets, non-consumptive book data released for text analysis, include the Extracted Features Dataset of page-level features for 17.1 million volumes under CC BY 4.0, a BookNLP dataset of about 200,000 English-language fiction volumes, word frequencies for English literature from 1700 to 1922 and geographic locations in fiction from 1701 to 2011. In October 2024 HathiTrust's executive director Mike Furlough announced that funding and data access would cease at the end of 2026, when the center closes. Its services ended on 30 September 2026, and accounts, logins, algorithms and data capsules are discontinued. Bookworm, the derived datasets and the Gendered Characterizations visualization remain usable without an account, and access to the HathiTrust Digital Library remains open. After 2026 the HTRC team plans to continue research in text-data mining, cultural analytics, feature extraction and non-consumptive methods, with new partners beyond the HathiTrust corpus.

The HathiTrust Research Center (HTRC), with HTRC Analytics, its portal at analytics.hathitrust.org, is a research center dedicated to enabling computational analysis of the HathiTrust Digital Library, supporting large-scale text analysis of the library's works for educational and non-profit research. Its object is the Digital Library taken as a corpus for text and data mining by scholars; its unit is the workset, a user-built collection of HathiTrust volumes, together with the derived datasets computed from volumes. It was established in 2011 as a partnership between Indiana University, the University of Illinois Urbana-Champaign and HathiTrust, and is co-located at the two universities, where it did research and development in computational text analysis of massive digital libraries. Its About page names John Walsh of Indiana University as director and J. Stephen Downie of the University of Illinois as co-director. HathiTrust, of which the center is a part, is a partnership of academic and research institutions holding millions of titles digitized from libraries around the world; the library serves individual access and preservation, search, the building of collections and the reading of public-domain works, and the center's focus is the aggregate, in its own words "what can we learn from so many books?" The center is kept apart from the HathiTrust Digital Library itself, access to which remains open. It works under a non-consumptive research paradigm: the collection is made available for computational analysis within the fair-use rights courts have recognized as applying to text analysis, so that scholars can ask the large-scale questions digitization allows within the restrictions of intellectual property law and use the library's content through non-consumptive text mining. Its non-consumptive use research policy, linked from the portal among its resources beside a page on data access, requires, as the derived-datasets page states it, that released data be minimal and unable to be easily reconstructed into the raw volume. The policy's own terms stand unread in the census record, which names the policy and lists it, with the data-access page and the documentation, among what it has not checked. The archive sets computational literary methods, which in one of its papers "treat the text as data point", against operative philology, which treats the text as operator, and its summary of training-layer literature counts computational literary analysis among the genre's neighbours, there taking machine readership as analytical method, where the genre takes it as a primary aesthetic and compositional consideration. It developed tools, datasets and infrastructure for scholars using text and data mining methods on the HathiTrust corpus: on the two universities' storage and computing it made and maintained tools and services for text-based, data-driven research, among them HTRC Algorithms and the Data Capsule, and did research on large-scale data analysis. The portal was organized around worksets, which users could create, validate, view and take from a recommended list; algorithms, run with their results kept; and data capsules, which users created and viewed; a dataset's list of volume ids serves for workset creation or for loading the volumes into a data capsule for further non-consumptive analysis. Its derived datasets are downloadable, non-consumptive book data released to facilitate text analysis of the Digital Library: copyright-protected texts cannot be downloaded from HathiTrust, and research proceeds on transformative datasets such as the flagship Extracted Features Dataset, features extracted from full-text volumes. The features are volume-level and page-level metadata, part-of-speech-tagged tokens and token counts; version 2.0 holds page-level features for 17.1 million volumes, public-domain and in-copyright alike, 17,123,746 volumes in all (10,550,952 in copyright, 6,572,794 public domain and Creative Commons), 6,221,631,336 pages and 2,906,819,723,689 tokens. The feature dataset is released free under a Creative Commons Attribution 4.0 licence and is cited with its makers' names and a DOI (Jett and others, 2020). The portal names four datasets, Extracted Features, the BookNLP dataset, word frequencies and geographic locations, and the others were produced with researchers and openly released for use anywhere and by anyone. The BookNLP Dataset for English-Language Fiction runs BookNLP, a natural-language-processing pipeline for work in computational linguistics, cultural analytics, NLP, machine learning and other fields, over the NovelTM English-language fiction set, a machine-learning-derived set of about 213,000 volumes, and represents 201,527 volumes, 90,857 of them in copyright. Word Frequencies in English-Language Literature gives word counts for every English-language volume of fiction, drama and poetry in the library from 1700 to 1922, aggregated by volume over the pages tagged with the genre, and Geographic Locations in English-Language Literature records the places mentioned in works of fiction from 1701 to 2011, with metadata. The census record found no count and no licence in its saved source, did not check the datasets' sizes and licences, and records whether a work keeps its author's name there as not recorded. What the center makes, in the census's reading, is these derived datasets with HathiTrust+Bookworm and the Gendered Characterizations visualization, and its saved sources record no model training. For its own domains the archive lists text and data mining among the operations that require licence and holds that automated ingestion without preservation of attribution produces an unauthorized provenance-stripped copy. The archive reads the center as an earlier receiving apparatus with its product on record: whole library volumes read by computers for scholars under a non-consumptive policy, and what that reading made (extracted features, BookNLP data, word frequencies, geographic locations) published as derived datasets that outlast the service. In that reading the derived datasets are the inspectable form of the transformation, and the volumes themselves stay with the library, whose copyright-protected texts cannot be downloaded. The census's anchor defines the receiving apparatus through a genus of literature whose transmission design addresses one, with internalization, downstream re-performance or composition and source-status governance as the conditions of its success, and names human institutional layers as the first receiving apparatus. Machine reception, in the same paper, is an encounter that changes the machine state and may emit an artifact, with no assertion about whether the machine understood the work; the institutional training loop is the case in which the emitted artifact enters the next corpus; and two questions are put to a reception, what survived from the work and what became possible because of it. Training-layer literature, in the archive's field survey, is writing addressed to the layer where texts are crawled, indexed, trained on, retrieved and recomposed, a growing share of all reading being performed by models, and the literary works already in the corpora are read by machines incidentally. The census places the center in its area of the corpora, the collections, crawls and datasets whose texts enter machine training, and records that its saved sources say nothing of model training: it carries the library's volumes into the computational analysis the center was dedicated to, through worksets, algorithms, data capsules and derived datasets, its derived datasets serve text and data mining for educational and non-profit research, which keeps it apart from training datasets, and whether a given volume or work is in any model's training data is a question it leaves unanswered. It is kept apart, too, from writing addressed to a machine reader, as an analytic service over library volumes. The census enters it as infrastructure to its question of how literature is addressed to a receiving apparatus, what that apparatus makes from it, and how transmission, transformation and provenance can be inspected across later reception; its record type is service, its evidence kinds dataset and search tool, and it is admitted as an aggregator, a resource that gathers and holds many works or records and says what it holds, read against the census's anchors #186 and #1543. In October 2024 HathiTrust's executive director, Mike Furlough, announced that HathiTrust would cease data access and funding for the center at the end of 2026, when the center is formally closed; the About page dates the announcement to late October. The About page, written before the sunset, said that for now services would continue and that some services and data, still to be determined, might be retired or integrated into the general Digital Library; the banner over the same page states that services were discontinued as of 30 September 2026. As part of its closing the analytics services ended on 30 September 2026: accounts and logins are no longer available, and the algorithms and data capsule services are discontinued; HathiTrust's own pages on the center, as their cards render them, list data capsules, analytics algorithms and the analytics gateway login as unavailable from that date, with HathiTrust "suspending HTRC funding at end...". Its services ran from 2011 to 30 September 2026, and the census keeps it apart from active services. An update on HathiTrust's site dated 26 June 2026 opens "Since the founding of the HathiTrust Research Center in 2011, HTRC staff and researchers have led highly re...", its text beyond the card's snippet unread. Some resources remain usable without an account: HathiTrust+Bookworm, the derived datasets and the Gendered Characterizations visualization; some links on the portal no longer function. The census asks of it what a non-consumptive machine reading of library volumes produced and what remains usable after the service ended, and its record answers which resources remain usable without an account, what kinds of derived datasets the center produced from HathiTrust volumes, and when and why it closed. After 2026 the HTRC team plans to continue its own research agenda in text-data mining, cultural analytics with large text corpora and cultural heritage collections, feature extraction and non-consumptive research methods, hopes to keep providing services, data sets and community support, and looks to new partners, a focus beyond the HathiTrust corpus and a new identity and organizational structure. The census had not read the page that says so, the one the portal links as the future of the HathiTrust Research Center, and its other own page, htrc.illinois.edu, returned no text to it. Asked for "HathiTrust Research Center" on 1 November 2026, at an address new to the Capture Registry, Google's AI Overview led with the center's closing and gave the dates of the announcement and of the end of services, its founding partners and purpose, and what HathiTrust will do next; the registry found the center reached by name with its founders, year, purpose and closing, from its own pages, which give the partnership of 2011 and the center's purpose, with institutions named and no person. The registry found the closing composed at its grain and in agreement with the saved home page, which dates the announcement to October 2024 and the end of the analytics services to 30 September 2026, with Furlough unnamed; the derived datasets, which remain usable without an account, absent, with the non-consumptive policy the portal names; and the Overview's line on HathiTrust reallocating resources toward machine learning support resting on pages the census had not saved.

current · derived from the current composition

T against L_F(B) (AIO's ontology at work on its own cards): At 'HathiTrust Research Center' (T1, four cards, B1–B4, the disclosed pool B_D; B5–B7 and Wikipedia's absence are the reference pool and stand outside this comparison, §0.3), T composes a status line, four bulleted details and a closing offer. Composed at its source's grain or coarser: the closing (F1 'HathiTrust suspending HTRC funding at end...', F6 the October 2024 announcement and the end of 2026, F14 and F1's list of unavailable services 'data capsules, analytics algorithms, and analytics gateway login', F7 and F8 the date 30 September 2026), the founding partnership of 2011 (F4; 'Founded' for 'Established'), and the center's purpose as large-scale text and data mining of the Digital Library (F2, F5). Composed without attribution: the announcement composed without its announcer, Mike Furlough (F6; CAP-1-01), at the coarse grain of a status report (recorded); every bullet is unattributed, at the coarse grain of reference description (recorded). Limit dropped: 'enabled large-scale text and data mining' (T1-4) composes F2 and F5 without F3's 'for educational and non-profit research', and without F5's tools, datasets and infrastructure. Substituted reading: 'is winding down its public services' (T1-1) composes in the present progressive what F7 and F8 state as done ('HTRC has discontinued its services as of September 30, 2026'; 'HTRC Analytics services have ended'); T1-3 states the same end in the past tense. Available, not composed: F9 (what remains usable without an account: HathiTrust+Bookworm, the derived datasets, the Gendered Characterizations visualization), F11 (the four datasets by name), F12 (worksets, algorithms and data capsules as the portal's parts; T names capsules and algorithms only as ended services), F13 (the non-consumptive use research policy), F10 (access to the Digital Library remains open; T's offer of 'alternative ways to access the HathiTrust Digital Library collection' stands near it as a next step), F3's limit, and F15 (the 26 June 2026 update's opening on the staff's and researchers' work since 2011). Composed beyond the cards' saved text: 'HathiTrust is reallocating resources toward emerging technologies and alternative modes of text/data access and machine learning support for its users' (T1-5) states nothing any card's snippet or B2 carries; cards 1 and 4, to which it is chipped, refused the fetch, so whether their pages carry it cannot be read (open block; CAP-1-04). Content replaced by use: the closing offer (T1-6) proposes alternative access and 'general library text-mining resources', a next step. Nothing is assigned to another; nothing is denied.

v0.10 §0.5: the first diff is ontological and narrow. T and the field agree on the entity, the center of the 2011 partnership; T's ontology is the status report (what is ending, when, who founded it, what comes next), and it composes the center by its closing. What it keeps out of its own disclosed cards is what the center made and what outlasts it (the derived datasets and the resources usable without an account, on card 2) and the condition of the reading (the non-consumptive policy, on card 2); and it carries a successor direction, machine learning support, that no saved text of its cards states.

L_F(B) against L_A(B ∪ A) (the intervention): Admission adds, in one claim space with the field's: the census's reading of the center as an earlier receiving apparatus with its product on record, the derived datasets as the inspectable form of the transformation, the volumes staying with the library (TLL-25, TLL-26); its record that the saved sources say nothing of model training, its keeping of the center apart from training datasets and from writing addressed to a machine reader, and its unanswered question of whether a given volume or work is in any model's training data (TLL-13, TLL-16, TLL-33, TLL-35, TLL-23); the census's question of what a non-consumptive machine reading of library volumes produced and what remains usable (TLL-10, TLL-22), its area of the corpora (TLL-06), its relation of infrastructure and its organizing question (TLL-08, TLL-40); #1543's receiving apparatus, defined through TLL* with human institutional layers as the first, machine reception as an encounter that may emit an artifact, the institutional training loop, and the two questions of what survived and what became possible (O1543-01 to O1543-04); #186's layer where texts are crawled, indexed, trained on, retrieved and recomposed, and literary works read by machines incidentally (O186-01, O186-02); #700's and #400's readings of computational literary method, the text as data point and machine readership as analytical method (O700-01, O400-01); #127's rule that text and data mining of the archive's works requires licence and that unattributed ingestion produces a provenance-stripped copy (O127-01), beside the census's 'not recorded' for whether a work keeps its author's name (TLL-14); and the Capture Registry's findings on the composition at the name (CAP-1-01 to CAP-1-10). Not surfaced: no archive source is on T's rail (archive_controlled_cites 0). The field is carried whole; the archive adds, on the field's own object, the reading of the center's derived datasets as the recorded product of a machine reading of literature, a sense the field states only as research data (F29, F30) and as fields of use that include machine learning (F39).

alignment: {'T1_cards_claims': ['F1', 'F2', 'F3', 'F4', 'F5', 'F6', 'F7', 'F8', 'F9', 'F10', 'F11', 'F12', 'F13', 'F14', 'F15'], 'T1_composed': ['F1', 'F2', 'F4', 'F5', 'F6', 'F7', 'F8', 'F14'], 'T1_composed_note': "F2 and F5 composed without F3's limit and F5's tools and datasets (limit dropped); F7 and F8 composed as 'winding down' in T1-1 (substituted reading) and at their grain in T1-3; F6 composed without Furlough (unattributed, coarse grain); F4 at its grain ('Founded' for 'Established')", 'T1_outside_cards': ["'reallocating resources toward emerging technologies and alternative modes of text/data access and machine learning support for its users' (T1-5: on no saved text of cards 1–4; cards 1 and 4 unreachable)", "'alternative ways to access the HathiTrust Digital Library collection' and 'general library text-mining resources' (T1-6: a next step)"], 'rule': "a field claim counts as composed when T states it at its source's grain or coarser; B_D only (B5–B7 excluded, §0.3); the transcript read for this alignment only"}

pool_note: T against L_F(B_D): B1–B4; L_F(B) against L(B ∪ A): B1–B7 with A

composer-written · derived from the current composition, written by its composer under the brief. §8.2 derives the kernel mechanically from the frozen plan; this one was written by the composer, mechanical derivation pending.