chrullis/relweave-business-data

relweave business relation extraction data

Passages from English Wikipedia articles about Nordic and European companies, labelled with typed entities (every mention, including pronouns and descriptions such as "the company") and typed relations under a fixed business schema of 6 entity types and 21 relation types. It is the data the models relweave-4b-base and relweave-1.7b-base were trained and evaluated on, for the relweave library.

Splits

split	chunks	articles	relations	entities	labels
train	10,611 (3,359 whole chunks + 7,252 windows)	670	55,092	76,596	10,111 completed by a second labelling pass; 500 list-rich chunks labelled by correcting model drafts
validation	208	30	1,468	1,981	completed
test	105	29	918	1,038	completed and audited relation by relation
test_fresh	64	13	434	615	one labelling pass

No article and no chunk text is shared between splits. Training articles that name three or more people or companies of an evaluation article, or its subject, were excluded. The exclusion was not complete: 75 of 532 person and company names in the test set (64 of 306 in the fresh test) also occur in training labels, and 10 (24) relation triples recur by name; the fresh test overlaps training on Heineken and Bouygues (review of 2026-10-08). Chunks are cut at sentence boundaries, median about 900 characters. Of the 10,611 training rows, 3,359 are whole chunks and 7,252 are overlapping windows of a few sentences cut from them (ids such as abb#1w2, median about 400 characters), with the labels restricted to what the window states; they were extra training examples for the generator. Drop them ("w" not in chunk_id.split("#")[1]) for one row per passage. The evaluation splits hold whole chunks only.

from datasets import load_dataset
ds = load_dataset("chrullis/relweave-business-data")              # train, validation, test, test_fresh
sources = load_dataset("chrullis/relweave-business-data", "sources")["train"]

Fields

field	contents
chunk_id	article id and chunk number, e.g. abb#1
doc_id	article id (one per Wikipedia article)
title, url	the source article
text	the passage
entities	list of {id, type, name, attributes, mentions}; mentions are the surface strings as written in text
relations	list of {type, source, target, attributes, modality, evidence}; source and target are entity ids; modality is asserted, negated, hedged or reported; evidence is the sentence that states it (empty in some correction rounds)
labels	how the chunk was labelled: completed, completed_audited, corrected_draft or single_pass

schema.json holds the schema (types, definitions and allowed endpoints). Relation attributes (dates, shares, roles) are recorded where the text states them but are not part of the scored output of the models.

How the labels were made

Labels were annotated by a large language model following written annotation conventions (the schema definitions below plus rules such as: every member of a list gets its own relation; a subsidiary is linked only to its nearest stated parent; a board chair is EXECUTIVE_OF, an ordinary seat BOARD_MEMBER_OF), then spot-checked. The first training labels were completed by a second pass after an audit showed the first pass missed relations; the test labels were audited relation by relation; the 500 list-rich training chunks were labelled by correcting the model's own drafts. Corrected labels agree with independent from-scratch labels at 0.876 typed F1, and two independent from-scratch passes agree at about 0.88, so expect some missing and some disputable relations. Methodology: docs/methodology.md in the relweave repository.

Schema

Entity types (6)

type	definition
PERSON	A human being.
ORG	A company, institution, family acting as owner, or other organised body; publications, imprints and brands are Orgs.
OBJECT	A physical thing: vehicle, vessel, building, artwork.
PLACE	A country, region, city or other location.
COORDINATE	A numeric map coordinate.
EVENT	Something that happened at a time: a sale, meeting, election, ceremony.

Relation types (21)

Edge attributes such as dates, shares and roles exist in the classes but are not part of the scored output. Symmetric relations have no direction.

relation	endpoints (source -> target)	definition
EMPLOYED_BY	Person -> Org	Person works for Org in a non-executive role
EXECUTIVE_OF	Person -> Org	Person holds an executive title at Org (CEO, CFO, managing director) or chairs its board or supervisory board (title chairman, vice chairman, honorary chairman)
BOARD_MEMBER_OF	Person -> Org	Person sits on the board of Org as a member; a board chair is EXECUTIVE_OF instead
MEMBER_OF	Person -> Org	Person is a member of Org (club, party, association)
FAMILY_OF	Person -> Person (symmetric)	Family tie; kind is what the target is to the source: spouse, parent, child or sibling
ASSOCIATE_OF	Person -> Person (symmetric)	Persons described as associates, partners or close contacts
MET_WITH	Person -> Person (symmetric)	Two persons met in person
COMMUNICATED_WITH	Person -> Person (symmetric)	Two persons communicated (call, email, message, letter)
LOCATED_IN	Object or Org or Person or Place -> Place	Source is or was resident or situated in the Place (dated if the text says when), or a Place lies within another; not a birthplace, not a market; an Org that is based in a Place is HEADQUARTERED_IN
BORN_IN	Person -> Place	Person was born in the Place
OPERATES_IN	Org -> Place	Org has operations, offices, plants, stores or sales in the Place
HEADQUARTERED_IN	Org -> Place	Org has its headquarters in, or is based in, the Place
FOUNDED	Org or Person -> Org	Person or Org founded or co-founded the Org; when Orgs merge to form a new Org, each merging Org FOUNDED it
OWNS_STAKE_IN	Org or Person -> Org	Person or Org owns shares in, a stake in, or controls through ownership the Org; share is the stated fraction
SUBSIDIARY_OF	Org -> Org	Source Org is a subsidiary, division or brand of the target Org; publications, magazines, newspapers, imprints and product brands are Orgs linked this way, never Objects; in a chain of owners, link only to the nearest parent the text states
ACQUIRED	Org or Person -> Org	Person or Org bought or took over the target Org; only when the text names the buyer
HAS_COORDINATE	Place -> Coordinate	Place has the stated numeric coordinate
OWNS_OBJECT	Org or Person -> Object	Person or Org owns the Object (a physical thing: vehicle, vessel, building, artwork); brands and publications are Orgs
TRANSFERRED_OBJECT	Event -> Object	The Event transferred the Object (sale, delivery, seizure)
PARTICIPATED_IN	Org or Person -> Event	Person or Org took part in the Event; role is buyer, seller, host, attendee
HELD_AT	Event -> Place	The Event took place at the Place

Licence and attribution

The passages are from English Wikipedia and are licensed under CC BY-SA 4.0; the labels are derived from them and are released under the same licence. Attribution: Wikipedia contributors; each row carries its article's title and url, and the sources configuration lists all 742 articles. Reusing or redistributing this data, or data derived from it, requires the same attribution and the same licence (share-alike). The models trained on it are released separately under the Apache License 2.0.

The texts were fetched from Wikipedia between 26 September and 3 October 2026; article revisions were not recorded.

Personal data

The passages name real people (founders, executives, board members, owners) as they appear in Wikipedia. To have a passage removed, open a discussion on this repository.

Limitations

• English business Wikipedia only: cleaner than news, filings or reports.
• One annotator model under one set of conventions; labels contain misses, mostly in long coordinated lists.
• The fixed schema above; other relation types are not labelled even where the text states them.