17

I must be dumb, but before entering this site I have never seen the "Ngram" word. So what does it mean? I have seen several answers here cites Ngram stats which seems to be discombobulating me. Can anybody cast some light on this?

19

I must be dumb, but before entering this site I have never seen the 'Ngram' word.

I had never heard of Ngrams before I started frequenting ELU. There's nothing ignorant about being unfamiliar with Ngrams.

Ngrams can be a very useful tool when trying to figure out which words and phrases are used in English. They can help show when certain phrases entered into the vernacular, and when they fell out of favor. But they have their limitations.

For example, say I wanted to know which word was used more often to describe a person who was unable to speak: mute or dumb. It would be very hard to construct an Ngram to get that information. I could construct one like this one and then jump to the conclusion that dumb is used more often. As you can imagine, though, that conclusion is likely erroneous, because the word dumb has other meanings, as in, "I must be dumb, but before entering this site I have never seen the 'Ngram' word" (which may be grammatically correct, although I disagree with that statement).

Also, it's important to remember: Where is Google getting that data? From published works – that is, books, journal articles, and magazines. So, Ngrams will tell you how authors use phrases – but not how people talk in their homes, or at the corner bars, or in the shopping malls, or even in online forums like this one. There are probably more than a few phrases which show up in the Urban Dictionary but would return a flat or near-flat Ngram. Ngrams don't tell the whole story. As an example, searching for "noob" on Google returns over 50 million hits, but the word barely registers in Ngrams, and has not yet found its way into many dictionaries.

One last thing to be cautious about when evaluating Ngrams is the amount of usage. For example, I could look at this Ngram and draw the faulty conclusion that the term "facsimile machine" became common after 1980. The REAL story, of course, is that shortly after "facsimile machine" started to gain traction, it was quickly eclipsed by the more familiar "fax machine."

Are Ngrams useful? Yes. Do they tell the whole story? Not always. They are but one tool in the English learner's toolbox. As long as you're aware of their limitations, though, they can be a handy way to confirm or refute your hunches about how English is written over time.

  • Great answer! Ngrams are useful, but are not an answer in and of themselves. +1 for that and also for the interesting "facsimile machine" tidbit--I'd never really even wondered where "fax machine" originated, so you taught me something new today! – WendiKidd Feb 8 '13 at 23:07
  • 1
    It might be worth mentioning that n-gram is a more general term (for contiguous sequences of N words), and Google books Ngram Viewer is the name of a tool which happens to report data based on low-order n-grams. And although most users will continue to use Google's data and interface, other sources of n-gram data are available. For this reason, I try to avoid referring to Google's tool as simply n-grams or Ngrams. – snailcar Sep 15 '13 at 15:42
  • For those interested in knowing how to overcome Ngrams' shortcomings, one trick is to use a phrase that is unlikely to be ambiguous, in the case of dumb vs mute, I'd suggest changing the subject to children. Calling children dumb because of foolish or irresponsible actions is unlikely to be used. books.google.com/ngrams/… – Mari-Lou A Sep 6 at 6:57
7

Most simply, Ngram charts show how often words and phrases are used in books over time, and often compared to other words or phrases. For example, you can check how common "double digits" is compared to "double figures".

You can also check different languages (technically, "corpora"), or compare them. US and UK English can be compared, so we can see if "double digits" is more common in the UK.

An "n-gram" is a word or phrase. The n refers to the number of words (or in some cases, word parts). "hat" is a 1-gram, "double digits" is a 2-gram. But usually this distinction doesn't matter to us.

See About Ngram Viewer for a lot more information and examples.

3

Others have commented on what an Ngram is.

What an Ngram is too often used for in language-related SE sites, unfortunately, is to add pretty and official-looking pictures to answers. I feel that, especially on an English Language Learners site, we need to be very careful about the use of such tools because frequency alone can be misleading to new speakers without other crucial context.

For example, Ngrams don't reveal if a construction is common but stigmatized or otherwise socially marked. Also, Ngrams have limitations with showing all the forms a construction might take.

0

I believe Wikipedia article is a good start.

It is widely used for linguistics, but not only. For instance, noise reduction algorithms are based on this phenomenon.

See "nlp" tags on Linguistics, SO, and Programmers. Theoretical Computer Science and Computational Science sites may be useful as well.

See also Markov chain.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .