NCIER®

Methodology

How the claims in this map were extracted from the source report, checked against its wording and highlighted on the original pages.

The map at a glance

This map covers Preventing Targeted Workplace Violence: A U.S. Secret Service Guide to Behavioral Threat Assessment and Management, published by the U.S. Secret Service National Threat Assessment Center in 2026. It was generated on 26 October 2026 and contains:

  • 1,382 claims
  • 23 sections
  • 100 concepts
  • 4,798 links between them
  • 100 source pages

A claim is one statement the report makes, written as a short title and a plain sentence, and tied to the exact words of the report that support it. A section is one of the report's own divisions. A concept is a recurring idea, such as a term or topic, that gathers every claim mentioning it. A link connects a claim to its section, to a concept it mentions, to another claim that supports it, or to another claim it cites.

How the claims were extracted

NCIER's research pipeline works through the source report page by page and records each claim it finds together with the passage of the report that supports it. Every claim carries the report page it comes from, a type, a title, a plain-language sentence, and the supporting quote. The pipeline runs offline; this website only publishes the finished dataset and never edits it.

Claims are classified into types:

How every quote is verified verbatim

The quote attached to a claim must be the report's own wording, not a paraphrase. Each quote is checked automatically against the text of the page it is attributed to, and the dataset has to pass that check for all 1,382 quotes: every one must be found on its page word for word.

Text taken from a PDF is not always typographically identical to the text typed into a quote, so both sides go through the same normalization before they are compared. Normalization changes only how characters are written, never which words are present:

  1. Curly single and double quotation marks are straightened to plain ' and " characters.
  2. Hyphen, non-breaking hyphen, figure dash, en dash, em dash, horizontal bar and minus sign are all treated as a plain hyphen.
  3. Non-breaking spaces become ordinary spaces.
  4. Table column separators (vertical bars) are replaced by a single space.
  5. Runs of spaces, tabs and line breaks are collapsed into a single space, and leading and trailing whitespace is trimmed.
  6. All letters are case-folded to lower case.

After normalization the quote must appear in the page text exactly. Reworded, shortened or reordered text does not pass. The check also confirms that every claim's section, concepts and supporting claims exist, so no link points nowhere.

Tags: verbatim-supported and inferred

Every claim carries one of two tags, shown in square brackets. The quote is always the report's exact wording; the tag says how closely the claim sentence follows it.

  • [V] Verbatim-supported (1,304 claims): the quoted words state the claim directly.
  • [I] Inferred (78 claims): the quote supports the claim, but the claim goes beyond the exact words. Every inferred claim states the reason, so a reader can judge the step for themselves. See all inferred claims and their reasons.

How page highlights were located

Each claim's quote was located on the image of its source page, and the highlight is stored as one or more rectangles measured as fractions of the page's width and height. The page images are the report's pages exactly as published; the highlights are drawn over them on this site and never baked into the image. 100 source pages are published at two sizes.

The way each highlight was located is recorded with the claim:

  • Exact (1,236 claims): the quoted words were found as one continuous passage on the page.
  • Table row (138 claims): the quote comes from a table, so the whole table row is highlighted.
  • Clauses (4 claims): the quote is made of separate clauses, and each clause is highlighted where it appears.

Claims with no highlight

For 4 of the 1,382 claims the highlight step found no match on the page image, so those 4 claims have no highlight. Their source is artwork, such as a figure or graphic, rather than running text, and the quote could not be tied to a rectangle. Their quotes are still verified against the page text like every other claim; they are shown without a highlighted page thumbnail.

  • p047-08 report p. 47: Figure 7 core members (top): Human Resources, EAP, Security, Ombudsman
  • p047-09 report p. 47: Figure 7 core member (left end): Management
  • p047-10 report p. 47: Figure 7 ad hoc participant (right end): Information Technology
  • p075-14 report p. 75: Diagram: off-ramp from Evaluate to Case Closure when appropriate

Use the data

Everything on these pages comes from one dataset, which you can download and check yourself.

Explore the graph Back to the map overview


Top

Find the Perfect Training Class For You