12

This is just a fun factoid to entertain you.

By now you should know not to put too much stock in Google Ngrams result graphs, and insist on a deep dive into the individual results. I thought I'd share a new kind of weirdness I encountered.

I had been curious as to which of the various spellings of McGuffin (referring to the plot device) was more common. My initial search yielded many results from the 19th Century.

I settled on this search, utilizing the definite article to eliminate persons named McGuffin or whatever. There were still 19th- and early 20th Century results! One was a Theodore McGuffin (abbreviated as The.) but even weirder things showed their heads.

Apparently, Google's OCR algorithm does not know how to recognize books printed with multiple columns on one page. So, in a two-word Ngram, you can get a result if one word in one column is near the other word in the other column, purely by coincidence:

mcguffin the

12
  • 2
    Yes, I can confirm that Google Ngram reads texts horizontally which can create very weird excerpts. Snippets don't improve the situation much.
    – Mari-Lou A
    Sep 4, 2021 at 19:02
  • 3
    NGrams is a great tool, but shouldn't be used simplistically. As @Mari-LouA has noted several times, don't just look at the graph, check actual instances found. Many of them. And even then be skeptical. But don't -stop- using NGrams because of its difficulties.
    – Mitch
    Sep 4, 2021 at 21:04
  • @Mitch My initial paragraph says exactly that.
    – Spencer
    Sep 4, 2021 at 21:35
  • 1
    @Spencer Your paragraph could be construed as saying "Don't use NGrams". It doesn't say it outright, but you only give the negative. I'm emphasizing that there are good things about NGrams, even though you should be careful how you use it.
    – Mitch
    Sep 4, 2021 at 22:49
  • Another hazard of working with Ngram (pointed out to me by Hugo at this site eight years ago, when I fell into exactly this trap) is that Google Books (and hence Ngram) tends to cite matches on the basis of the title page that appears nearest the front of a database selection—but some databases bundle multiple pamphlets as a single selection, so you need to scroll back from the match you find to page 1 to confirm that it is indeed the first page of the whole database selection, and not the start of a different pamphlet with a different title (and, often, a different date) in the same bundle.
    – Sven Yargs
    Sep 5, 2021 at 7:03
  • 1
    As a side note, it would probably be good to find somewhere to report that OCR issue to so it can potentially be fixed (even if Google is rather notorious for not responding to users).
    – NotThatGuy
    Sep 12, 2021 at 13:23
  • 2
    Perhaps some kind soul can tell me how to create a tag: it is high time we had a tag for spoken versus written English. NGRAMs can be informative for written English (and some plays and scripts) but it's pretty worthless as far as spoken English is concerned.
    – Lambie
    Sep 15, 2021 at 14:57
  • @Lambie you can do that already if you have a question of spoken x written English, you have enough rep - here
    – Juliana Karasawa Souza
    Sep 16, 2021 at 6:02
  • 1
    @JulianaKarasawaSouza Thanks for responding. Yes, I have read that link before and as I understand it, you have to ask a question to create a tag. But I have no question to ask. Why is everything so complicated around here? :)
    – Lambie
    Sep 16, 2021 at 13:30
  • @Lambie You can always edit the tags of an existing question, as long as the new tags are appropriate......
    – Spencer
    Sep 17, 2021 at 12:25
  • 1
    @Spencer Why is such a simple question I posed so hard to understand? I do not understand how to create a tag. The directions say: You have to ask a question to create a tag. I do not have a question I want to ask. I only want to create a tag. Why is this site so hard to use?
    – Lambie
    Sep 17, 2021 at 14:48
  • 2
    @Lambie Those directions are incomplete! Edit an existing question. Go down to the tags section. Type in the new tag name. Submit. If you have any problems, ask a new question on meta. Stop adding comments to this one.
    – Spencer
    Sep 17, 2021 at 23:16

0

Reset to default

You must log in to answer this question.

Browse other questions tagged .