Datasets:
Adversarial Audit: Proto-Indo-European (ine-pro)
Source Verification
- Claimed source: wiktionary
- Entry count: 841
- Entry count plausible: YES (expected 500-1000 for PIE)
Format Verification
- Header correct: YES (
Word\tIPA\tSCA\tSource\tConcept_ID\tCognate_Set_ID) - All rows have 6 fields: YES (0 malformed rows)
- Duplicate entries: 0
Content Verification (10 random samples)
| # | Word | IPA | SCA | Gloss | IPA Valid? | SCA Valid? |
|---|---|---|---|---|---|---|
| 1 | gʰedʰ- | ɡʱedʱ | GED | to join | YES | YES |
| 2 | ksnew- | ksnew | KSNEW | to scrape | YES | YES |
| 3 | kʷey- | kʷej | KEY | to pay | YES | YES |
| 4 | leykʷ- | lejkʷ | LEYK | to leave | YES | YES |
| 5 | seykʷ- | sejkʷ | SEYK | to moisten | YES | YES |
| 6 | (s)leh₃y- | (s)leɣʷj | SLEGY | blueish | YES | YES |
| 7 | stru- | stru | STRU | gray-haired | YES | YES |
| 8 | ne | ne | NE | not | YES | YES |
| 9 | sweks | sweks | SWEKS | six | YES | YES |
| 10 | sweḱruh₂ | swekruħ | SWEKRUH | mother-in-law | YES | YES |
Hallucination Check
- Round entry count: NO (841)
- Generic glosses found: 0
- Empty fields: 0
- Word==IPA entries: 3/841 (0.4%)
- Empty/missing concepts (dash): 0/841 (0%)
- Wiki artifact glosses: 0
Cognate & Source Metadata
- Cognate_Set_ID: all entries are "-" (not populated)
- SCA character inventory: A, B, D, E, G, H, I, K, L, M, N, O, P, R, S, T, U, W, Y (19 classes, valid)
Verdict: PASS
Notes
- This is the highest quality lexicon of the 6 audited. The 0.4% Word==IPA rate means virtually all entries underwent meaningful IPA conversion.
- Laryngeal handling is well-executed. PIE laryngeals (h₁, h₂, h₃) are converted to IPA equivalents: h₁ -> h, h₂ -> ħ (pharyngeal fricative), h₃ -> ɣʷ (labialized velar fricative). This is a linguistically reasonable mapping following Beekes/de Vaan conventions, though the exact phonetic values of PIE laryngeals remain debated. 151 of 841 entries contain laryngeal reflexes.
- Aspiration notation: Aspirated stops (bʰ, dʰ, gʰ, gʷʰ) are converted using the breathy voice diacritic (bʱ, dʱ, ɡʱ, ɡʷʱ), which is the standard IPA representation for voiced aspirates. This is phonetically precise.
- Labiovelar handling: kʷ, gʷ, gʷʰ are properly represented as labialized consonants with the IPA superscript ʷ.
- Palatovelar handling: Palatovelars (ḱ, ǵ, ǵʰ) are mapped to plain velars (k, ɡ, ɡʱ) in IPA, which is a defensible choice since PIE palatovelars are conventionally transcribed but their exact phonetic value is debated.
- The lexicon contains both verb roots (e.g., "bʰeg- to break", "gʷem- to come") and nominal/adjectival forms (e.g., "h₂stēr star", "pṓds foot"), which is the expected structure of a PIE reconstructed lexicon.
- All 841 entries have glosses, and the glosses are linguistically specific and consistent with standard PIE reconstructions (cf. Pokorny, Rix LIV, de Vaan). Some glosses reference academic sources (e.g., "", ""), which are Wiktionary citation artifacts but do not compromise data integrity.
- Entry count of 841 is well within the expected range and represents a substantial portion of the reconstructed PIE lexicon.