Understand the entities inside an article
Move from a draft to an editorial brief: identify the subjects, inspect their mentions and use the surrounding context to guide your next decision.
Based on official documentation · Reviewed September 14, 2026
What entity extraction does
Entity extraction turns article text into a structured list of its subjects: people, organizations, places and other recognizable things. It helps you see who and what a story is about, then connect those subjects back to the passages where they appear.
A mention is the actual phrase in the article. An entity is the subject that phrase refers to. Keeping the two connected matters when a person is named in several ways or the same word could mean different things. TextRazor’s entity reference describes entity identifiers, positions, relevance and confidence, alongside links to knowledge sources where available.
D2TR brings detected entities together with categories, entity trends, article context and available Discover matches. Each part of the analysis has its own status, so you can distinguish completed output from a stage that is still processing or unavailable.
Reading an analysis
Google opened an office in London.
- Entities and mentions
- Recognized subjects, available entity types and the number of mentions. Select an entity to connect it to the annotated article.
- Content categories
- Topic labels that describe the document. They classify the article as a whole rather than listing every person or organization named in it.
- Trends and context
- Available demand charts and a structured editorial brief help you explore the subject beyond the text. These signals need their own interpretation.
- Covers and similar posts
- Available reference images and related publication matches. Public previews retain the access boundaries shown in the workspace.
Google’s Natural Language documentation distinguishes entity analysis from content classification: one identifies subjects and their mentions, while the other assigns document categories. Neither output replaces a reader’s judgment of the article.
Analyze a draft
- Prepare the article. Edit the supplied draft or paste English text into the editor. The public workflow accepts 200–20,000 characters.
- Keep useful structure. Headings, paragraphs, lists and tables can be edited in the workspace. Include enough context to make the subjects unambiguous.
- Run Analyze text. Watch the analysis progress, then inspect the annotated article and the completed result stages.
- Review and refine. Select entities, inspect categories and explore available trends or publication matches. Switch back to Article editor when you want to revise the draft and run a new analysis.
The request inspector above lets you explore this sequence on the page: submit a document, receive a job ID and access token, then read the job until it completes. A job can contain useful successful stages even when another stage fails or returns reduced data. Check each status before using its findings in your review.
When to use it
Review the focus of a draft
Compare the intended story with the subjects that the analysis detects. Repeated names and highlighted passages can reveal where the article spends its attention. Use that evidence to support an editorial discussion about clarity and focus.
Prepare topic and entity research
Start from the people, companies and places mentioned in a draft, then inspect their context. When an ID is available, verify the subject with Knowledge Graph Finder before adding it to a research list.
Improve metadata consistency
Use extracted subjects and categories as candidates for tags, desk labels or a content inventory. Review suggestions before applying them, especially when a name is ambiguous or the document covers several unrelated themes.
Prepare an editorial handoff
Run an analysis on this page before sharing a draft with an editor. Record the subjects and findings worth reviewing alongside the relevant version of the article, so the next reviewer can connect each observation to its source text.
Keep the evidence in context
More mentions do not automatically mean more importance. For example, Google defines entity salience as how central a subject is to the document, rather than simply how often it appears. Its entity field reference explains this distinction. Do not treat different providers’ scores as interchangeable.
Discover context is useful for research, but it does not establish that a draft will appear in Discover. Google says indexed content that meets its policies is automatically eligible, while eligibility does not guarantee an appearance. Consult Google’s Discover guidance when making publishing decisions.
For the best review, read the underlying article alongside the results. Check uncertain names, open available source links and treat unavailable stages as missing evidence rather than a negative judgment about the draft.