Select the statement that best describes the organization today, one per
chapter. The result shows a level for each chapter and the chapters to read
first. Answers are stored in this browser only.
1. Define How is use of the data defined and named?Not yet Use means downloads; datasets are referred to by whatever name a document uses. Foundational A written definition distinguishes mention from use; each dataset has one canonical name and acronym. AI-ready A typology of use (primary, secondary, background) and a canonical names table with variants and identifiers are maintained. AI-native The names table is the one the catalog and the recommended citations use, and it is updated from the monitoring results. 2. Access How are downloads and API calls counted?Not yet From raw server hits, or not at all. Foundational Every access path is logged with dataset, action, client, user agent, and status, and counted monthly with robots and double-clicks excluded. AI-ready Counts follow the COUNTER Code of Practice for Research Data with regular and machine access separate, kept as a series per dataset. AI-native Usage is reported to DataCite for datasets with DOIs, machine access is attributed to registered integrations, and counts feed prioritization. 3. Citations How are formal citations counted?Not yet Datasets have no DOIs; citations are not counted. Foundational The main datasets have DOIs and a citation format; citations are looked up by hand. AI-ready Citation events are collected from DataCite, Crossref, and OpenAlex on a schedule and merged per dataset. AI-native Identifier citations are compared with text-mined mentions each run, and one-source documents feed the collection and the DOI backlog. 4. Collect (research) How are documents that may use the data collected?Not yet They are not collected; mentions come to the organization's attention by chance. Foundational A few sources are searched by hand each quarter. AI-ready Listed sources are collected on a schedule with document metadata (year, type, source) and a text layer. AI-native Collection runs automatically within the terms of each source, with coverage measured against a reference list. 5. Collect (web) How are policy, news, and web documents collected?Not yet By hand, when someone notices a mention. Foundational Government sites and news archives are searched by hand each quarter for the canonical names. AI-ready Document lists, feeds, and exports are collected on a schedule; crawling runs within robots.txt, identified and rate-limited; news and web datasets add breadth; coverage is measured. AI-native Collection runs automatically with the robots check as a gate, and coverage per source type is tracked over time. 6. Detect How are mentions found in documents?Not yet By reading, or by searching for the exact dataset name. Foundational Keyword search for the canonical names and acronyms. AI-ready A named-entity extractor finds mentions, with precision and recall measured on a labelled sample. AI-native Extraction runs in a pipeline, thresholds are tuned on the sample, and the sample grows with each review. 7. Harmonize How are mentions matched to datasets?Not yet They are not; counts are by mention text. Foundational Exact matches to the canonical names and acronyms. AI-ready Variants, acronyms, and near matches are resolved to identifiers, with unmatched mentions reviewed and the table updated. AI-native Semantic matching and clustering resolve new variants, editions are distinguished, and other organizations' data are tracked as co-use. 8. Report How is use reported?Not yet Download counts, if anything. Foundational A list of documents that cite each dataset. AI-ready Use per dataset by year and document type, with access statistics and the coverage and recall caveats stated. AI-native Reports include AI systems' answers, trends over time, and are published with the data. 9. Act How do the results change what the organization does?Not yet They do not. Foundational Results inform occasional decisions about which products to continue. AI-ready Documentation and dissemination work is prioritized by use; misquotes are followed up. AI-native Document records and series pages link to each other from the results, and corrections flow back to the catalog. 0 of 9 answered Show reading path Reset
Interpretation of the result
Chapter 1 first: without a canonical names table, detection and
harmonization measure noise.
Chapters 2 and 3 are the cheapest evidence: logging and DOIs take days
and the counts arrive every month after that.
Chapters 4 to 6 can start small; a few sources and a labelled sample of
a hundred sentences give the first real numbers.
Chapter 9 is where the value is realized; a report that changes nothing is
a cost.