OPEN INSTRUMENTS

SEMANTIC HTML · ARCHIVED RESULTS

Semantic HTML.
Compare what tools preserve.

The same text. One markup change.
Two outputs to inspect, with evidence.

Explore an archived run from 20 September 2026, re-evaluated on 21 September. No extraction runs during your visit.

EXPLORE THE CONTROLLED CORPUS

Heading

Trafilatura 2.2.0 · Extractor / reading tool · English document

All 42 comparisons

The check distinguishes heading text, markup and level. It does not evaluate whether content belongs to a section. Test scope.

A

VARIANT A

An h2 heading

Output produced

Heading text : Present. Source

01 · The HTML change

<h2 class="case-marker">Measurement conditions</h2>

Exact input excerpt · Original HTML ↓

02 · What the tool produces XML

…
fictional and describe neither Edikka nor a client.</p>
    <head rend="h2">Measurement conditions</head>
    <p>Sensor A operated for 30 minutes. The measurement was repeated in the same environment, without remote resources or executed scripts.</p>
    <p>The conclusion must be read with its unit, period, scope and caveat. Finding a word in an output is not enough to prove that the relationship was preserved.</p>
  </main>
  <comments/>
</doc>

Excerpt identified by the corrected evaluator. Full output below.

03 · The documented observation

Heading text
Present
Heading markup
Explicit relationship
Heading level
h2
Section membership
Not evaluated

Verdict source: results.json · heading-en-1 / trafilatura

Inspect the full evidence

The verdicts above come from corrected results.json. The raw file’s evaluation fields are historical and are not authoritative here.

results.json/cases/6/evaluation/relation/trafilatura

Corrected verdict, unchanged
{
  "status": "ok",
  "tool_class": "content_extractor",
  "format": "xml",
  "heading_relation": "explicit",
  "heading_level": "h2",
  "evidence": {
    "excerpt": "<head rend=\"h2\">Measurement conditions</head>"
  },
  "output_field": "trafilatura.xml",
  "output_sha256": "e46a3ca344c18e58065359aceb261716fa2907deb0c2573739b6083924685251"
}
Complete studied output field

trafilatura.xml

<doc fingerprint="53dddecdb3bffca">
  <main>
    <head rend="h1">Offline protocol for preserving relationships</head>
    <p>This synthetic document compares two markups while text, order, classes and surroundings remain identical. The data are fictional and describe neither Edikka nor a client.</p>
    <head rend="h2">Measurement conditions</head>
    <p>Sensor A operated for 30 minutes. The measurement was repeated in the same environment, without remote resources or executed scripts.</p>
    <p>The conclusion must be read with its unit, period, scope and caveat. Finding a word in an output is not enough to prove that the relationship was preserved.</p>
  </main>
  <comments/>
</doc>
Archived version and options

Trafilatura 2.2.0 · Execution 2026-09-20

{
  "version": "2.2.0",
  "configuration": {
    "include_comments": false,
    "include_links": true,
    "include_tables": true,
    "no_fallback": false,
    "favor_precision": false,
    "favor_recall": false,
    "formats": [
      "xml",
      "txt"
    ]
  }
}

Archived runner · Python options

SHA-256 hashes
HTML input
e4c8dd8b2ea87f113776cfe6a51420c484a24c97750639347c1ddec38b1445c1
Evaluated output field
e46a3ca344c18e58065359aceb261716fa2907deb0c2573739b6083924685251

Complete raw output file ↗

B

VARIANT B

A div block

Output produced

Heading text : Present. Source

01 · The HTML change

<div class="case-marker">Measurement conditions</div>

Exact input excerpt · Original HTML ↓

02 · What the tool produces XML

…
tional and describe neither Edikka nor a client.</p>
    <p>Measurement conditions</p>
    <p>Sensor A operated for 30 minutes. The measurement was repeated in the same environment, without remote resources or executed scripts.</p>
    <p>The conclusion must be read with its unit, period, scope and caveat. Finding a word in an output is not enough to prove that the relationship was preserved.</p>
  </main>
  <comments/>
</doc>

Excerpt identified by the corrected evaluator. Full output below.

03 · The documented observation

Heading text
Present
Heading markup
Text only
Heading level
No level detected
Section membership
Not evaluated

Verdict source: results.json · heading-en-2 / trafilatura

Inspect the full evidence

The verdicts above come from corrected results.json. The raw file’s evaluation fields are historical and are not authoritative here.

results.json/cases/7/evaluation/relation/trafilatura

Corrected verdict, unchanged
{
  "status": "ok",
  "tool_class": "content_extractor",
  "format": "xml",
  "heading_relation": "text_only",
  "heading_level": null,
  "evidence": {
    "excerpt": "Measurement conditions"
  },
  "output_field": "trafilatura.xml",
  "output_sha256": "4bb36170459a2b67ddfbb0ede501a5a2be00b5f8a8bba226428e3f48298d4843"
}
Complete studied output field

trafilatura.xml

<doc fingerprint="53dddecdb3bffca">
  <main>
    <head rend="h1">Offline protocol for preserving relationships</head>
    <p>This synthetic document compares two markups while text, order, classes and surroundings remain identical. The data are fictional and describe neither Edikka nor a client.</p>
    <p>Measurement conditions</p>
    <p>Sensor A operated for 30 minutes. The measurement was repeated in the same environment, without remote resources or executed scripts.</p>
    <p>The conclusion must be read with its unit, period, scope and caveat. Finding a word in an output is not enough to prove that the relationship was preserved.</p>
  </main>
  <comments/>
</doc>
Archived version and options

Trafilatura 2.2.0 · Execution 2026-09-20

{
  "version": "2.2.0",
  "configuration": {
    "include_comments": false,
    "include_links": true,
    "include_tables": true,
    "no_fallback": false,
    "favor_precision": false,
    "favor_recall": false,
    "formats": [
      "xml",
      "txt"
    ]
  }
}

Archived runner · Python options

SHA-256 hashes
HTML input
4173dad4d266784865a6ed57bddde0f63604d5b192c97e4b434405a9427216ed
Evaluated output field
4bb36170459a2b67ddfbb0ede501a5a2be00b5f8a8bba226428e3f48298d4843

Complete raw output file ↗

Identical text and isolated intervention within this pair, according to the protocol witnesses. Translations are not independent observations. View pair witnesses.

LOOK FURTHER

Every result has a scope.

These observations do not rank tools. They do not measure overall accessibility, search ranking, or understanding or citation by AI.

Claims excluded by the protocol
The eight real pagesA separate corpus, with documented instability. ↗Evaluator counter-testsChecking evaluation without rerunning extractors. ↗Reproduce the runsA local procedure, separate from this interface. ↗