Talk:Word n-gram language model

	This article has not yet been rated on Wikipedia's content assessment scale. It is of interest to the following WikiProjects:
Lists This article is within the scope of WikiProject Lists, an attempt to structure and organize all list pages on Wikipedia. If you wish to help, please visit the project page, where you can join the project and/or contribute to the discussion. ??? This article has not yet received a rating on the project's importance scale.

Todos

Latest comment: 2 years ago2 comments1 person in discussion

• Add a history section. Jurafsky and Martin has a short but useful section about this.
• Add a section on smoothing
• Add a section on applications

Colin M (talk) 17:25, 10 March 2023 (UTC)Reply

I was also thinking of merging in a bunch of content from n-gram (which is currently an awkward combination of being about n-grams themselves and n-gram models). But there's a complication in that that article covers n-gram models as applied to a broader range of sequences, where as this article is currently focused on modelling sequences of words. We could have yet another article about n-gram models more broadly, but it doesn't seem like there are enough differences to make that distinction worth it. Probably better to broaden the scope of this article to match that. Colin M (talk) 17:56, 10 March 2023 (UTC)Reply

N-grams have not been superseded

Latest comment: 1 day ago1 comment1 person in discussion

Related to the above comment, the article begins with the claim that n-gram models have been "superseded" by neural models. This is nonsense, as the latter are completely different. N-grams model information in a local window, while the cited neural models (and presumably most others) carry information over an arbitrary distance. If one wanted to model local sequential structure in language, such as many phonotactic restrictions (restrictions on adjacent sound sequences) then n-grams remain appropriate and valuable. Khanson679 (talk) 02:01, 26 August 2025 (UTC)Reply