Sources & Method

How the archive is built

This page explains what is in the corpus, how records are summarized and scored, and where the limits are. The explorer is meant to speed up review, not replace the original source files.

Corpus

Public source material

The archive is built from public release files and DVIDS-hosted videos. Individual records keep source links so claims can be checked against the original PDF or video page.

Workflow

Extraction pipeline

  1. Parse release documents into record-level entries.
  2. Extract dates, locations, agencies, record types, tags, and source URLs.
  3. Pair videos to related PDF records when the IDs and context match.
  4. Render the data as browsable JSON for the static site.

Scoring

How rankings should be read

Useful triage, not official truth.

What raises a case

Operational detail, multiple objects, sensor references, speed estimates, unusual morphology, observer assessment, and strong source linkage can increase the anomaly score.

What the score is not

It is not a government finding, scientific proof, or a confidence rating that the event is unexplained. It is a sorting signal to decide what deserves closer reading.

How to cite a record

Open the source PDF or DVIDS page, confirm the extracted summary, then cite the original source rather than the score or generated synopsis alone.

Release Ledger

What is included

Limits

Caveats that matter

OCR and extraction

Older scans, redactions, tables, captions, and low-resolution imagery can create imperfect text. The source file remains the authority.

Locations

Map positions are approximate and derived from release locations, theatre labels, or operational context. They should not be read as precise coordinates.

Completeness

The site reflects the release data present in this build. Empty or zero-count release entries are preserved to show the state of the source corpus.

Source Files

Build inputs