Talk:Heaps' law

[Untitled]
The original version of this page was adapted from http://planetmath.org/?method=src&from=objects&id=3431&op=getobj owned by akrowne, with permission under the GFDL

Divergence
I think with "Where VR is the subset of the vocabulary V represented by the instance text of size n" the author wanted to say "Where VR is the cardinality of the subset of the vocabulary V represented by the instance text of size n", because a subset is not a number. However, the size of the subset diverges (i.e. becomes arbitrarily large) as n goes towards infinity. That would only make sense if the vocabulary would also be of infinite size. (just as a sidenote: I would have expected the fraction of the vocabulary not covered by the text to decrease exponentially when looking at larger and larger documents). Icek (talk) 19:02, 29 September 2009 (UTC)

Vocabulary size is infinite according to generative grammar, see e.g. Mark Aronoff "Word formation in generative Grammar" MIT Press 1985, Andras Kornai "How many words are there?" Glottometrics 2002/4 61-86 88.132.28.96 (talk) 20:33, 17 March 2012 (UTC)

Types, Tokens, and Hapaxes: A New Heap's Law
There is a new paper by Victor Davis (link), deriving a stronger version of Heap's law from first principles (rather than empirically). I think it's worth adding it to the article. Compare this YouTube video for a brief summary and visual explanation. Renerpho (talk) 20:13, 29 August 2022 (UTC)
 * ArXiv preprints do not count as reliably published sources for Wikipedia purposes. YouTube videos even less. —David Eppstein (talk) 20:50, 29 August 2022 (UTC)
 * Fair enough -- How about this version? Renerpho (talk) 11:31, 30 August 2022 (UTC) The YouTube video is not intended as a source, but as assistance for the editor who is going to work on this. Renerpho (talk) 11:35, 30 August 2022 (UTC)