8

I am writing a paper and part of my thesis, and I need some advice to improve the readability of my manuscript(s).

Working in computer science and machine learning, there is a lot of technical stuff going on: I have to write definitions and set the terminology of my work. So, for instance, I have to explain the reader that a training set is a set of example on which my system learns to model some phenomenon.

Now, please, focus on the previous paragraph. As you can see, I used the italics to stress the new piece of terminology, namely the expression “training set”, and to emphasize some other words, namely the word “learn”.

So, having the same style for these two things, with a different semantics, sounds a little confusing to me (and to my supervisors). And there is a lot of this stuff in my paper/thesis.

How can I improve readability in this respect?

--
EDIT. The focus of my question is not if the word "learn" should be emphasized with respect to what it means or not. It is just an example. Idem for the "training set" with respect to the expected background knowledge of the reader. In abstract terms, my question can be generalized to: how to emphasize both the terminology that I am introducing and words that need proper emphasis? A more fitting example could be the following:

We call training set the set of examples on which a Neural Network learns to model some phenomenon. (omissis) At each timestep, our network has to decide if the symbol will be chosen from the first or the second list. In both cases, it has to decide also which symbol will be chosen.

  • 5
    I'm in a different area but I've seen only new terms emphasized when they are defined. Emphasizing words like "learn" just for the sake of it doesn't seem natural in academic texts; have you seen that elsewhere? – JiK Jan 10 '18 at 10:05
  • 1
    Just as a head's up, SE.TechnicalCommunication looks like it may be entering private beta soon, if you're interested in stuff like this. – Nat Jan 10 '18 at 10:17
  • 8
    Can you please edit your question to specify why you emphasise learn in said sentence? You said that it’s for a different reason but you do not specify which. – Wrzlprmft Jan 10 '18 at 10:19
  • 4
    @Nat Looks like a bad idea to me, conflating two very different targets (documentation and scientific papers). I wish I could downvote Area51 proposals, but apparently the SE folks are just interested in finding out how many more users it can attract, not in whether it is a good idea. :( – Federico Poloni Jan 10 '18 at 10:33
  • 1
    “[It] sounds a little confusing to me (and to my supervisors).” — Weird. Why is that? It’s pretty much the norm, and not generally seen as confusing. – Konrad Rudolph Jan 10 '18 at 14:11
18

Perhaps you don't need to italicise the word learn: you could just say "a training set is used to train the system to model the particular phenomenon".

This means that the italics are reserved for the words or phrases that you are setting the precise meaning of.

  • 3
    Agreed. Develop a style for definitions, where it is immediately clear that you are now defining a new term (e.g. by a big Definition 2.5 above) and decide on a style for the new term to be defined; e.g. bold, italic, etc. Then stick with that and do not use it elsewhere. If you want italic for other emphasis, then better take bold for new terms. If bold is also preoccupied, you need to put some more thought. – Dirk Jan 10 '18 at 10:44
  • 1
    Thanks for your reply. The "learn" think is just a excuse to explain. A more realistic scenario could be the following: We call training set the set of examples on which a Neural Network learns to model some phenomenon. (omissis) At each timestep, our network has to decide if the symbol will be chosen from the first or the second list. In both cases, it has to decide also which symbol will be chosen. I will edit the original question. – petrux Jan 10 '18 at 14:13
  • 6
    @petrux The meaning is completely clear without italicizing those words for emphasis. – Morgan Rodgers Jan 10 '18 at 19:21
  • For what it's worth, "our network has to decide if the symbol will be chosen from the first or the second list" is ambiguous to me. This could mean (1) or (2), where: (1) the network has to make a choice of whether the symbol will be chosen from one of these two lists [this further bifurcates according as to whether "exactly one" or "at least one" is meant] or the symbol is chosen from one or more other lists; (2) the network has to make a choice of exactly one list from among the first list and the second list. – Dave L Renfro Jan 11 '18 at 7:06
14

First of all, your university or target journal may have a style guide which specifies how they want such things to be typeset. So check whether they do, and if so, follow it.

My personal preference is to use boldface for terms which are being defined, and italics for general emphasis. This is because it is pretty common for a reader to need to refer back to a definition, and it is helpful if they can glance at a page and immediately spot the term they are looking for. Boldface is eye-catching and can be immediately picked out from a page or block of text, so it's good for that.

When you just want to emphasize a word, you want it to be noticed by someone who is already reading that sentence, but it isn't so important that they be able to glance at the page and jump immediately to that word. In your example, somebody might be skimming the page looking for where you have defined training set, but nobody is going to be skimming for the word "learn" which just happens to be important in that sentence. So italics is fine in that case.

As illustration, you might look at the following two paragraphs. In which one is it easier to spot the specific word?

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Curabitur in luctus dolor. Pellentesque orci ante, pulvinar ac leo ut, venenatis commodo tortor. Proin faucibus tincidunt nisl, quis semper tellus ullamcorper ac. Mauris quis tellus eleifend, condimentum diam id, posuere eros. Pellentesque habitant morbi tristique senectus et netus et malesuada fames ac turpis egestas. Donec bibendum auctor tristique. Donec justo odio, commodo id elit vitae, efficitur dictum massa. Phasellus placerat convallis ipsum et fermentum. Pellentesque urna mi, malesuada sagittis vehicula sit amet, auctor sed leo.

Phasellus ante quam, convallis eget vestibulum id, rutrum sit amet libero. Duis ultricies ornare semper. Nunc cursus aliquam ultricies. Morbi tellus neque, euismod sit amet tincidunt eu, volutpat interdum massa. Nunc eu elementum massa, sit amet vestibulum metus. Quisque hendrerit nunc feugiat ligula elementum, in mollis felis aliquam. Praesent hendrerit dolor id nulla lobortis fringilla. Aenean non dictum risus, id molestie lectus. Donec id ullamcorper nunc. Proin eu risus vitae leo hendrerit luctus. In scelerisque, purus scelerisque convallis imperdiet, leo nibh tempor ante, at ultrices ligula diam a dui. Ut eu risus id justo vulputate sagittis eget et tellus

  • This is exactly the approach I've come to prefer, but there are certainly publishers who will change all your boldface to italics. – Mark Meckes Jan 10 '18 at 17:36
  • 1
    Boldface does work well for terms being defined but I've never seen a pure mathematics or theoretical computer science journal use anything other than italics. And I think the intended purpose of the typesetting is to make it clear exactly what term is being defined. If I write "A proper colouring is..." then it's not clear whether I'm defining "colouring" and contrasting it with something that I feel to be improper or defining "proper colouring". But "A proper colouring is..." is unambiguous. You'd find the definition by scanning for Definition 7. – David Richerby Jan 10 '18 at 19:31
  • 3
    @DavidRicherby: I'm in pure math, and I've been able to use boldface for definitions in a number of my published papers. Example (needs subscription). The problem with Definition 7 is when you see "proper coloring" elsewhere in the paper, you don't know which definition number to look for. Also, I often find it nicer to define a number of terms in a prose paragraph instead of giving each one a numbered definition. – Nate Eldredge Jan 10 '18 at 20:40
7

Typographical emphasis is typically used for the following semantically different aspects:

  • introducing a new term
  • titles of works
  • referring to a word/term as such
  • species names
  • stress
  • highlighting an in-paragraph heading
  • person names
  • phrases from another language in English

As you typically only have limited number of types of typographical emphasis available (boldface, italics, small caps, …), you inevitably have to overload at least one of them in any larger work, usually italics. However, in most works and cases, it is easy to deduce from context which of the above cases applies. While this may not appeal to the structure-loving scientist in you, this kind of overloading is a common feature of human communication – and compensated by redundancy. Consider for example the different meanings of the word that in the English language.

That being said, if you really think your texts benefits from a clear distinction, you have to get creative in terms of emphasis. For example, some philosophical works use a zoo of quotation marks to distinguish different types of referring to a term. However, going by your topic, you almost certainly shouldn’t do something like this as it would only make your work more difficult to read.

  • Thanks for the suggestion. I would like to "keep it simple" too. – petrux Jan 10 '18 at 14:18
  • 1
    I think small caps for introducing technical terms is somewhat common – sgf Jan 10 '18 at 14:22
  • @sgf I think so too, and like it, but if you have a lot of acronyms/initialisms they'll be in caps or small caps which makes it less clear. – Chris H Jan 10 '18 at 15:01
4

I agree with @Solarmike's answer, but on a more general point of view, if this is a concern for you then it may be the case that you over-use italic for emphasis. For instance, I realized in the past that I tended to over-use quotation marks and parentheses: since then I try to avoid them whenever I can, and I think my prose has improved. When you re-read your text, try to go over each instance and see if you can remove it.

2

I’m with Solar Mike that learns doesn’t need to be italicized. Still, the cleanest solution seems to me to be adding a glossary section somewhere. You should certainly have the space in a thesis. If you can’t do it in a paper, you could still have a paragraph in the introduction along the lines of

In this paper, we use training set to refer to ___ ...

and from that point on, never italicize training set. Finally, you could try using quotation mark to define the terms instead of italics.

2

You're italicising training set and learn (sticking to your original example) for the same reaon -- to draw attention to them. So using the same emphasis for both is fine. If you have a long section on terminology, perhaps much of your introduction chapter is devoted to this, you can try paragraph headings:

Training set

We call training set the set of examples on which a Neural Network learns to model some phenomenon. (omissis) At each timestep, our network has to decide if the symbol will be chosen from the first or the second list. In both cases, it has to decide also which symbol will be chosen.

The choice of training set is of particular importance because...

Test data

The test data set is another dataset. Can you tell I'm making this up as I go along?

But I suggest doing something slightly different: instead of putting the term you're introducing as a paragraph heading, draw attention to it in a margin note. This acts almost as as index to the terminology section. Here's an example from Ohanian -- physics (an undergrad textbook I keep on the shelf):

enter image description here

This is all in addition to any glossary, as this terminology section will contain much longer descriptions and will be sorted in a logical rather than alphabetical order

  • It's a nice suggestion in a book, but I suspect margin notes will not be allowed by most journals / proceedings, especially if they are going to be physically printed (printed journals don't like to waste paper and so aren't likely to have wide enough margins to contain a margin note). And for a thesis, it's pretty common for a university to have strict formatting guidelines, including a requirement that margins be of a particular width and empty (so that text is not cut off when converting to microfilm or other formats). – Nate Eldredge Jan 10 '18 at 15:35
  • "to draw attention to them": No. He emphasizes "training set" because this is the notion being defined. It is common to italicize words that are being defined, and this is semantically different from merely stressing the word. – darij grinberg Jan 10 '18 at 15:49
  • @NateEldredge I was concentrating on the thesis aspect. Here in the UK guidelines don't tend to be all that strict so the option would be there – Chris H Jan 10 '18 at 15:49
  • @darijgrinberg if it was structured like a dictionary, I would agree with you, but in this case it's just emphasis; the reason for the emphasis is the novelty of the term – Chris H Jan 10 '18 at 15:51
  • Here is an example of confusion: "While a presheaf cannot generally be expected to [...], a sheaf will always have this property.". Is this a definition of a sheaf, or just a property that holds for all sheaves but perhaps for some presheaves as well? The only way to tell is by checking whether sheaves have already been defined before. – darij grinberg Jan 10 '18 at 15:53
2

Improve readability by not over-using italics for emphasis.

Text gets difficult and very annoying to read if the author keeps emphasizing words that they feel to be important. Especially if their emphasis doesn't quite agree with where you'd put it (you'd probably find it more natural if I'd emphasized "very", rather than "annoying", for example).

If you look at well-written, well-edited text, you'll see that typesetting is almost never used for emphasis. Technical and formal writing isn't intended to mimic the spoken word and it's better to use sentence structure or vocabulary to give the emphasis you need. For instance, you could describe something as "crucial" instead of "important" or "very important".

In the example you give in the question, you don't need any emphasis: "learns" is a verb and we naturally see verbs as important parts of sentences. If you feel more emphasis is needed, then contrast the learning of the training set with the use of other data: "The training set is a set of examples from which the system learns to classify data, as distinct from the whatever-you-call-it data that it will be deployed to classify." (I've kept "training set" italicized, since it's standard in computer science to italicize terms at the point where they're defined.)

0

Expect your reader to have some knowledge. Somebody who reads a new paper about a machine learning topic hopefully knows what a training set is.

Reduce the points that are basic knowledge for the target audience and your problem should solve itself.

0

Other options are bolding, different font, quote marks, a colon, or, if you really want to set it apart, a separate line.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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