Entity Prominence Map

An entity prominence map shows how consistently your key named entities — people, organisations, products, concepts — appear across the sections of your document. Entity salience is grounded in published NLP research: mention frequency, first-mention position, and distribution across document sections predict how dominant an entity is in a text (Dunietz & Gillick, EACL 2014; GUM-SAGE, ACL 2025). This tool computes a composite salience score from these three signals and maps each entity's coverage across paragraphs. User-specified terms are tracked directly. When fewer than five terms are provided, the tool supplements with the most frequently recurring multi-word phrases in the text.

The heat map shows which paragraphs contain each of your tracked entities. A filled cell means the entity appears in that paragraph; an empty cell is a gap. The goal is not 100% coverage — not every paragraph needs to mention every entity — but the final third of your document should reinforce your key terms, not abandon them. That is where coverage gaps matter most.

The bar chart shows what percentage of paragraphs contain each entity. An entity appearing in under 30% of paragraphs is effectively a one-off mention regardless of how many times it appears in total — it is not distributed across the document.

If the auto-detected entities are not the terms you care about — for example, if your key phrases are lowercase or not capitalised mid-sentence — use the "Track specific entities" field to specify them directly.

Entity salience — the degree to which a named entity dominates a document — is an established concept in NLP research and is used by information retrieval systems to assess topical focus. Content where key entities appear consistently across sections is more likely to be treated as authoritative on those entities than content that mentions them only once. This tool applies that principle to estimate entity prominence. The direct link between document-level prominence and AI citation frequency is a reasonable extrapolation from entity salience research, not a directly measured outcome.