Skip to projects Skip to filters

CORDIS · data.europa.eu competition

SDG mapping dashboard

Unofficial prototype

Download CSV

Projects

Loading mapping…

↑↓ move · → or Enter open evidence · ← or Escape close

What the mapping shows

How labels are produced

  1. Read CORDIS title, objective, keywords and EuroSciVoc titles.
  2. Compare that text with the official EU Vocabularies SKOS wording of each of the 169 SDG targets (http://data.europa.eu/sdg/{code}), using skos:broader for the parent goal. Indicator-level concepts exist in the taxonomy; they are statistical series, so we do not assign them to projects.
  3. Keep a target only if distinctive terms overlap. Isolated keywords are not enough. Goals that Horizon jargon would otherwise flood (1, 5, 7, 9, 12, 17) need extra anchors (poverty, gender equality, renewable energy, infrastructure, circular waste, partnerships).
  4. EuroSciVoc may boost an already supported goal; it never invents one.
  5. The scorer is sklearn TfidfVectorizer plus cosine similarity. No commercial LLM is called at inference. Weak CSV rows are the uncertain bucket the brief asks for.
Limits

Abstracts are promotional. Pure basic research may correctly receive no SDG. Treat uncertain (CSV: weak) rows as curator hypotheses, not catalogue facts. Scores are not an impact evaluation of a grant.

SDG concept

CORDIS abstract