Can corpus statistics tell whether a small symbol corpus like the Pictish one is language?
Most likely: the published measures do not reliably separate writing from non-linguistic systems; better measures exist…Ahead on the evidence, not yet confirmed. It needs two reviews to be accepted.
Where it stands
Four answers have been proposed. The published measures do not reliably separate writing from non-linguistic systems; better measures exist… leads, with evidence 2 for: partly supported by its tests. Three are still possible.
Why it matters
The Pictish corpus is under a thousand symbols in one- to three-symbol texts, which strains any statistical test[1]. The answer decides how much weight the 2010 and 2014 results can bear.
Answers 4
History
- Lee et al. puts forward: a calibrated entropy decision tree can classify small symbol corpora by type, and it shows the Pictish…
- Lee et al.: the authors' decision tree separated their own calibration corpora of known writing and non-linguistic…For: A calibrated entropy decision tree can classify small symbol corpora by type, and it shows the Pictish…
- Fournet: linearising carvings into strings and archaic-script features may bias such tests.For: The published measures do not reliably separate writing from non-linguistic systems; better measures exist…
- Sproat puts forward: the published measures do not reliably separate writing from non-linguistic systems; better measures exist…
- Sproat: none of the previously proposed methods were useful as published on the larger comparison set.For: The published measures do not reliably separate writing from non-linguistic systems; better measures exist…
- Asked on this site