0

The specific context here is Machine Learning technical writing, but I think this question fits here because I just haven't seen any discussion of this anywhere, and few examples in my field, so I'm not 100% sure a "standard" already exists.

If I have a model that is trained on some dataset (e.g. ImageNet), it is very common to call that an "ImageNet-trained model," as in the title of this conference paper. However, many datasets have numerals in their official titles (e.g. TMDB 5000). My question is:

If I were to reference a model trained on such a dataset, is the proper way to hyphenate that:

  1. "TMDB-5000-trained"
  2. "TMDB 5000-trained"

I have referred to many style guides (e.g.) and while they somewhat lead me to believe that option 1 is correct, I can't find good examples of it anywhere in practice to confirm that, even in ML writing. (This is largely because most ML datasets already use a hyphen in their title, such as "ImageNet-11k." I'm instead curious about nouns like "TMDB 5000," which are somewhat less common.)

2
  • 1
    Generally, the hyphen like that would be adjectival, as is the case with that conference paper. So, personally, I'd not use the second hyphen.
    – Lambie
    Commented Oct 26, 2022 at 17:00
  • "model trained on X"
    – Stuart F
    Commented Oct 26, 2022 at 21:58

1 Answer 1

2

If you're going to hyphenate, then you should include hyphens between all the words in the adjectival phrase. The fact that these are technical terms or include hyphens doesn't impact it. Consider other examples:

  • I made my peanut-butter-loving kid a PB & J. (Adds one word, loving, to a common two-word phrase peanut butter, to function as a three-word adjectival modifying kid.)
  • Peter had to dodge crowds of Spider-Man-obsessed girls. (Takes a hyphenated proper noun, Spider-Man, and adds a word.)
  • I've had it with your I'm-hungry-for-dessert-but-not-for-brussels-sprouts attitude! (Yup, the whole thing still modifies attitude.)

Disclaimers:

  1. Usage is not a matter of morally-charged correctness or incorrectness; if you're not operating under a defined style guide, you can do as you please, as long as you communicate your intent.
  2. Hyphenation is often omitted in these cases, so you could say "We're looking for a TMDB 5000 trained specialist" and probably be understood, and certainly not universally scolded.
  3. But I would argue that the hyphens help communication. It's all too easy to read the sentence from left to right and think you're looking for a TMDB 5000 unit, and then wonder what a "trained specialist" refers to. Even if the reader figures it out by the end of the sentence, they hit a roadblock of confusion in the middle.
2
  • 1
    I agree with everything, but I would argue that your "I would argue" sentiment is important enough that it belongs in its own point, not in parentheses. It's a way to tie your two numbered points together: this is a matter of style, and it matters only because, or only when hyphenation would make the writing more clear.
    – Juhasz
    Commented Oct 26, 2022 at 20:40
  • 1
    @Andy Your third point here is exactly what prompted me to ask this! I figured it out, but I got confused reading a sentence in a report and wanted to suggest a revision. Thank you!
    – SRobProsc
    Commented Oct 27, 2022 at 18:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.