France Occupation Direct Crosswalk v1

What it is

All 1,911 job profiles (fiches) of France Travail's ROME 4.0 (version 61) are linked directly, not via ISCO-08, to an occupation master of 1,033 nodes. The nodes are the 1,016 O*NET-SOC codes in the O*NET® 31.0 Database (O*NET-SOC 2019 taxonomy) and 17 France-specific "other occupations" (ACW-FR-NNNN) for which AI Changing Work found no O*NET-SOC counterpart. Relations follow the definitions of the ESCO–O*NET crosswalk (2022) and always read from the O*NET-SOC side.

Three AI judges worked independently: two Anthropic Claude runs using different evidence, and one OpenAI Codex run. A link is confirmed when the third judge took part and a majority held, provisional when it rests on the two Claude runs (or on a rule applied to them) without the third judge, and reviewed when the operator checked it. Most links (2,314 of 3,418) are provisional.

This is AI Changing Work's judgment mapping. It is not a correspondence from France Travail or from USDOL/ETA, and France Travail has not reviewed it.

• The links do not pass through ISCO-08 (International Standard Classification of Occupations). The occupation descriptions of the two classifications were compared directly.
• France-specific occupations judged to have no O*NET-SOC counterpart were not forced into a match; they were added to the master as other occupations (ACW-FR-NNNN).
• Relation definitions follow the technical report of the ESCO–O*NET crosswalk (European Commission × U.S. Department of Labor, 2022), and relations always read from the O*NET-SOC occupation side.
• Used together with the Korea Occupation Direct Crosswalk, which links the Korean classification to the same O*NET-SOC code backbone, it connects the two countries' classifications through the same master. The Korean other occupations (ACW-KR-*) are not included in this package.

Key figures

Occupation master

1,033

1,016 O*NET-SOC codes + 17 other occupations (France)

ROME fiches linked

1,911 / 1,911

0 fiches left unlinked

Links

3,418

match 1,809 · related only 1,609

Status

1,025 · 2,314 · 79

confirmed · provisional · reviewed

Provisional links have not been cross-checked by a second model (Codex). To use confirmed links only, filter the status column to confirmed and reviewed.

Of the 1,016 O*NET-SOC occupations, 966 have a ROME link (830 of them a match) and 50 have none. Of the ROME fiches, 1,539 have a match (including the 17 listed as other occupations). ROME is more finely divided than the O*NET-SOC taxonomy in many places, so onet_broader is common.

The English and Korean names are AI Changing Work's translations of the French ROME titles, not names set by France Travail (paired masculine/feminine forms reduced to one). The "nearest O*NET-SOC occupation" is the same as the related links of that fiche; it is not a match.

Code scheme: ACW (publisher, AI Changing Work) · FR (ISO 3166-1 country code) · NNNN (serial number in order of listing). The serial number carries no meaning, and once assigned it is never changed or reused. The source code (ROME) is kept separately in the source_code column. This is the same scheme as the Korean other occupations (ACW-KR-).

How it was made

1. Forward mapping — 923 O*NET-SOC occupations → ROME fiches (3 representative fiches per occupation, plus further fiches from the same split). Links from step 1 that disagreed with the ESCO path were judged once more with the same judge setup, adding 6 links (step 1′).
2. Relation review — the Korean table's step of this name does not exist in this table. The numbering follows the Korean table.
3. Orphan reverse mapping — the 641 fiches that no O*NET-SOC occupation pointed to in step 1 → the 1,016 O*NET-SOC occupations. The 17 fiches for which a majority of judges found no fitting occupation were added as other occupations ACW-FR-0001 to 0017
4. Review — the operator checked orphan reverse-mapping links without judge agreement and the other-occupation listings

Step 1 covered 923 = 1,016 O*NET-SOC occupations − 74 "All Other" residual occupations − 19 military (SOC 55); these 93 were reopened as candidates in step 3. Step 3 is the same step as the "orphan validation" of the ESCO–O*NET crosswalk. The forward mapping starts from the O*NET-SOC side, so some fiches are never chosen (641 of 1,911, 33.5%). Judging them from the opposite direction connects every fiche to some node of the master.

• The three judges did not see each other's answers.
• The ROME material given to the judges was the French source text of the France Travail CSV files, and the O*NET material was the English source text.
• The judges were given, as candidate hints, the fiches reached mechanically along the ESCO path (O*NET-SOC code → ESCO–O*NET crosswalk → ESCO↔ROME job-title correspondence V1.4 → version 61 fiche). The hints carried fiche codes and titles and the ROME job titles (in French) passed through, with no relation labels, and the judges were instructed to search beyond the hints. Because the correspondence is from an older edition, fiches created in 2024–2026 are rarely among the hints. The hint material is not included in this table.
• X was assigned to rows where A and B disagreed. In step 3 it judged all such rows (370); in step 1 it judged only some (including a sample of 5 rows where A and B agreed). In step 1, disagreeing rows without an X judgment were resolved by the operator rule below.
• The published files contain codes and titles only. They do not contain ROME definition or skill statements, O*NET definition or task statements, or the judges' reasons. ROME titles are copied unchanged from libelle_rome in the version 61 CSV (including paired masculine/feminine forms).

78% of the step-1 forward-mapping links (1,805 of 2,302) are provisional. They have no cross-model (Codex) check, so they are less reliable than confirmed links.

• Operator rule (2026-09-28, for step-1 rows where A and B disagreed and X did not judge): for a fiche both chose, take the weaker relation (related < close < narrower/broader < exact; if narrower and broader split, close) · for a fiche only one chose, adopt it if it is on the ESCO hint path (split), otherwise reject it. The ESCO path was also a hint in the judges' input, so it is not independent evidence. All 958 links that entered through this rule are provisional.
• The "further fiches from the same split" that judges listed beyond the 3 representative ones (22 links) were recorded as onet_broader, as the brief defines.
• Human review: the 74 step-3 links where the three judges disagreed or partly agreed were reviewed and approved one by one by the AI Changing Work operator on a side-by-side source comparison screen (2026-09-30). The 17 other occupations are fiches that a majority of judges found to have no corresponding O*NET-SOC occupation; 5 of them, where A and B agreed and X did not judge, were listed as reviewed after the operator checked them (no objection, 2026-09-30).
• Rule checks: 0 violations of "at most one exact per O*NET-SOC occupation". 0 of the 1,911 ROME fiches left unlinked.

Limitations

• This is a judgment mapping. It has not been validated against employment statistics. o_in_r and r_in_o (the judges' estimate, made before choosing a relation, of what share of workers in one occupation fall within the other side) are the judges' estimates, not measured headcount ratios.
• The judges are AI models. Human review was limited to step-3 links without agreement and to the other-occupation listings. The step-1 operator rule is not a row-by-row review.
• Provisional links are 68% of the total (2,314 of 3,418). Agreement between two judges of the same model family is not strong evidence (in the Korean table, Codex gave a different relation for 34% of the relations on which the two Claude judges agreed).
• The ESCO hint path chains an older-edition correspondence (from the ESCO × ROME V3 period), so it does not preserve relations, and because it was the basis for rule adoption (split), it is not independent evidence for those links.
• Military occupations (SOC 55) and the "All Other" residual occupations were left out of step 1 and were candidates only in step 3.

Changes

• v1 (2026-10-01): First release.

Attribution required when you share this data

AI Changing Work (2026). France Occupation Direct Crosswalk v1 (ROME 4.0 ↔ O*NET-SOC 2019). https://aichanging.work/data/fr-direct-crosswalk. Licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/).

Terms of use and sources

Below is the full text of the package's LICENSE file, reproduced as published. In short: AI Changing Work's contributions (links, relation labels, match class, stage, consensus and status values, and the ACW-FR-* nodes with their English and Korean titles) may be used under CC BY 4.0 to the extent AI Changing Work holds rights in them. O*NET-SOC codes and titles remain under the O*NET® 31.0 Database license (CC BY 4.0) with the attribution below, and ROME codes and titles remain under the Licence Ouverte 2.0 (Etalab); AI Changing Work does not relicense them.

The English and Korean names of the other occupations (ACW-FR-*) are AI Changing Work's translations, not names set by France Travail.

Download

These files are byte-identical to the published package. You can check a downloaded file against its SHA-256, which is also listed in CHECKSUMS.sha256.