Basics

1. The canonical order of thoughts in many languages is first given, then new. In other words, the first part of a paper, paragraph, or sentence should be what is given or well-known, the remaining part should be what is being presented as new. It is not easy to explain this rule in general, but consider a few examples.
   • My students like pizza. Given: that I have students; new: what they like.
   • What my students like is pizza. Given: that I have students and they like something; new: that what they like is pizza (and not gyros, say).
   • The ones who like pizza are my students. Given: some people like pizza; new: they are my students (and not Pete's students).
   For each such sentence, consider its complement to get a sense of what it is that the sentence emphasizes.
2. Use the active voice where possible (almost always).
3. Use a personal style with "you" and "we" instead of "one." For example, "you will find the server unacceptably slow" instead of "one will find the server unacceptably slow." You might add a little bit of formality by avoiding "you," but in that case you should write "users will find the server unacceptably slow."
4. Use "we" instead of "I" to refer to the authors of a paper. However, over-use of either personal pronoun can get tiresome, so replace it by neutral entities, e.g., "This paper develops a ..." instead of "We develop a ..." or "Section 3 shows ..." instead of "In Section 3, we show ..."
5. Use apostrophes for possessives [and genitives, if you want to be technical :-)], but not for plurals. That is, avoid constructions like "ETD's" when you mean multiple ETDs. Use "ETD's" in sentences of the following form "The ETD's link is broken."
6. Don't use plurals to describe distributive properties. For example, instead of saying "agents play roles" say "each agent plays a role" or "each agent plays exactly one role." A many-to-many relationship is much less informative than a one-to-one relationship (assuming the latter is what holds).
7. For any doubtful construction or clause, check if eliminating it would cost you anything. Text that can be deleted without loss of meaning should be. Often, the sentences that give us the most grief (in getting their phrasing right) are not quite as essential as we may initially believe.
8. Don't use any more pronouns than you need to: this, it, and others. They are often confusing. In particular, referring to something with a pronoun but after a gap of a sentence or two is highly confusing.
9. When you are citing works by others, use their names. Include a citation, of course. When the citation style requires a numeric reference to a bib entry, the names of the authors are even more valuable. When the citation style requires names and years, you can omit the names in the references. However, all the styles I can think of allow reducing the citation to just the year. So with numeric references, we end up with "Chopra [3] says ..." whereas with name-year references, we end up with "Chopra [2022] says ..."—these are both good solutions.
10. I see no benefit from saying things like "The authors of [7] show ..." or "In [7], the authors show that ..." In both cases, you end up using a lot of words without conveying a lot of meaning. Think also about the implicature of the above snippets. Could it possibly be anyone other than the authors of a publication who might have shown something through the publication? To me, the above constructs sound as bizarre as the famous real-estate advertisement or placard saying "for sale by owner"—sale by an owner is hardly newsworthy because it is after all the law. I would be much more impressed and informed by an advertisement that said "for sale by disgruntled neighbor"!
11. Don't use comparators unless you specify a comparison: "our approach is faster" [than what?] or "this problem has garnered less attention" [than what?].
12. Don't use superlatives as comparators: "our approach fastest than X".
13. Use "than" as in "higher than" instead of "higher as compared to".
14. Don't use meaningless statements that impugn another approach without clarity. For example, in "this problem has received limited atttention", it isn't clear what "limited" means. Everything
15. Use hyphens for word compounds "agent-based"; en-dashes for number ranges "10–15"; and em-dashes for certain kinds of parenthetical remarks. Here's an example using them all: "agents—well-known for at least 10–15 years—have become increasingly popular of late."
16. Don't use hyphens where the prefix is not a standalone word. For example, "re-do" is better written as "redo." The main exceptions are words where the removal of a hyphen would cause ambiguity. For example, "I resent your message" might be better stated as "I re-sent your message" unless you truly do resent the given message :-). I have occasionally encountered "resign" as well.
17. Use hyphens where compounds are used as modifiers, and not as plain nouns, e.g., "this is a run-time error" and "this error occurs at run time." Similarly, "this is a well-defined case" and "this case is well defined." Also watch out for "low level" and "low-level" and its cousins.
18. When you use a number range (or a range of other enumerations), it identifies one conceptual object covering that range. For example, 2008–2015 is a range of years. You may write "E occurred over 2008–2015" or "E occurred between 2008 and 2015" but not "E occurred between 2008–2015" or "E occurred from 2008–2015." There are few easier ways to rile me up :-).
19. Some phrases are naturally used both as phrases and as words run together. For example, "set up" and "setup" are both OK, but in different uses. You set up your experiment [verb]; what you will end up with is a setup [noun]. I would treat "feed back" and "feedback" the same way.
20. Don't use past or future tense for describing what is coming up in a paper. That is, avoid "We have presented a formalization of this hocus pocus in section 5" or "We will present a formalization of this hocus pocus in section 5." Instead, say "We present a formalization of this hocus pocus in section 5" or, better, "Section 5 presents a formalization of this hocus pocus." In other words, get rid of the "have" and the "will"—this is one of the few contexts in English where the simple present is acceptable and desirable.
21. Write "e.g.," and "i.e.," with commas as here. There would be some punctuation preceding the "e.g.," and "i.e.," snippets—this could be a comma or a semicolon most often.
22. Never begin a sentence with "E.g.," or "I.e.," or any other such abbreviation.
23. Don't use "however" to separate sentence-level clauses. For example, "I like pizza. However, it makes me sleepy." is OK as is "I like pizza. It makes me sleepy, however." but "I like pizza, however, it makes me sleepy." is not good.
24. Don't use "however" where "howsoever" is intended. "This agent can process sessions, howsoever many may be initiated." would be acceptable to indicate that the agent is not sensitive to the number of sessions. "Howsoever" is not used a whole lot any more, however.
25. "That is" and its common abbreviation "i.e." yield a clause of the same grammatical type as the clause that they elaborate. For example, "My friend won, i.e., got selected" is OK, but doesn't sound all that good. "My friend, i.e., buddy, won" is also OK. "My friend won; i.e., he got selected" is OK. Notice the distinction between the comma and the semicolon. It is not OK to write "my friend won; i.e., got selected," because a semicolon separates clauses that include a main verb. Likewise, it is not quite as OK to write "my friend won, i.e., he got selected."
26. As another distinction between commas and semicolons, consider sentences with so-called subordinate clauses. Commas help us separate such clauses. Instances of these are the relative clauses ("A, which is a B") as well as contrastives ("A is good, but B is bad"). You won't use semicolons in such cases.
27. Remember that when a subordinate clause is inserted parenthetically, it must be delimited by a comma on each end. It is as if open and close parentheses were replaced by commas in the text. It is a common error to forget the "closing" comma. For example, "A, which is a B sends a message" is bad; it should be "A, which is a B, sends a message." Likewise, we need "Both types of agent, consumers and providers, may initiate a conversation."
28. Look out for spurious commas. For example, "the commitment, the buyer makes" is bad; instead you could write "the commitment that the buyer makes" or "the commitment the buyer makes."
29. Learn to distinguish "that" from "which." The previous sentence was my longstanding admonition, but it hasn't worked, so here's more detail:
    • P that Q means {x: P(x) and Q(x)}. For example, "large computers that are fast" refers computers that are both large and fast.
    • P, which Q means {x: P(x)} and by the way P entails Q. For example, "large computers, which are fast" refers to large computers and further claims that large computers are fast. People who have heard of ENIAC and interpret "fast" in modern terms recognize this claim to be false.
30. Use "only" carefully. A slight modification to its placement in a sentence can drastically alter meaning. Consider "can only be achieved by ..." versus "can be achieved only by ..."
31. Don't switch perspective across sentences without appropriately warning the reader. For example, if one sentence talks about "fetching" a result, the next shouldn't say "returning" a result. Decide which side you are on.
32. Don't exaggerate. If you say there are several shortcomings in Smith's approach, you should list several—more than a few (to use another vague word). Otherwise, just say "some shortcomings" (still vague, but not too bad if you back up your claim with specific shortcomings) or "the following shortcomings" or something to that effect.
33. Don't precede an enumeration with an exact count, as in "the following three points ..." Often, the points will end up being split or combined. Or, some points will be deleted and new ones inserted. Such counts end up wrong unless you take special care to keep them consistent. Best to simplify one's task and not include a count in the first place.
34. Read your prose for how it sounds. You can end up with what I will call inadvertent phrases, which are amusing at best. For example, "an agent advertises a good deal"—does it mean a deal is advertised that is good or that the agent advertises excessively? Another example was in the CfP for a long-running series of conferences, whose boilerplate included the gem: "papers departing from guidelines or arriving late [will not be considered]"—you wonder if the departing and arriving are related like the departure and arrival of a train or bus are related.
35. Constructs such as "add additional slots" or "linear lines" or "choose among choices" or "we distinguish among three distinct separate points" sound careless at best, and unintentionally funny.
36. Don't say "on the other hand, ..." unless you had a previous sentence that began "on the one hand, ..." It is often better to use something like "conversely, ..." or "in contrast, ..." or "by contrast, ..."
37. Don't use longer constructs such as "there is a semantics for X" when you can instead use "X has a semantics."
38. Don't add noise words. For example, "this box is blue in color" could just as well be "this box is blue"—after all, it can't be—blue in length. Likewise, "business logic is proprietary in nature" could be "business logic is proprietary."
39. "Less number of" manages to pack two errors in three words
40. Use "it" (likewise, "its" and "their") for impersonal objects such as tables, chairs, organizations, agents. Use "he" and "she" for people. For animals, choose as you see fit.
41. Replace constructs such as "the following are X that are Y" by "the following X are Y." Save two words!
42. A construction such as "Each of X ..." means (for all x in X: ...). That is, X is a set—more formally, an expression whose extension is a discrete set. In English, we would write X in the plural form. There is an associated convention that you shouldn't use this construction if X has fewer than two members.
43. A construction such as "One of X ..." means (for some x in X: ...). The rest of the previous comment applies here too.

Articles, determiners, and definite descriptions

This is a complex subject and one that I couldn't possibly do justice to. Try to build your awareness of it. Here are some hints to help you begin.

A definite description seeks to specify particular objects (often, just one object). The objects may be abstract or concrete. A typical means in English is the use of the "the"—also, called the definite article. A more general idea is of determiners, examples of which include are "my agent," "the message's parameters," and "the third message."

The easy cases are where the presence or absence of a determiner leads to an ungrammatical construction. The harder cases are where it changes the meaning.

1. Over-specified definite descriptions are ungrammatical. For example, "the Figure" or "Figure 3" are OK, but "the Figure 3" is not.
2. "Literature" and "The literature" are both fine. The former evokes thoughts of Homer and Shakespeare, however, and is less likely to be of use in a computer science article.

Prepositions

Choose your prepositions carefully.

1. Ancestor of, not to.
2. Beside is not the same as besides: choose carefully.
3. Called, not called as.
4. Centered on, not around.
5. Combining together. Would you combine separately?
6. Compliance of.
7. Comply with, not to.
8. Comprises of: comprises would do as would consists of.
9. Consider, not consider about.
10. Cooperating together. Would agents cooperate separately?
11. Discuss about. Just discuss would do.
12. Emphasize, not ever emphasize on.
13. Focused on, not around (it wouldn't be focused then, would it?).
14. Normalize X to, not on.
15. "It seems that" can be as effective as "It seems to me that."

Words and Phrases to Avoid

Don't use the following words and abbreviations:

1. Slashes, as in "some scientists/researchers found/discovered/conjectured that computers/information processing systems are best modeled/understood/implemented as input/output automata." A slash is another one of those constructions that indicates a profound carelessness of thought. Does the slash indicate an and, an or, or an association? It is as if one person is writing as if he or she were a committee.
2. And/or: a particularly unhelpful (but common) special case of the above. The obvious question is "is it and or or"?
3. Pronouns: Avoid phrasings that rely on pronouns. A trouble with pronouns is that their referent isn't always as clear to the reader as it might be to the writer. If you can come up with names (for people, projects, approaches), you can simply write out the name each time.
4. His/her; he/she: I hate this construction with a passion. If you do need pronouns, you can say his or her and he or she. Or, you can use plurals instead of he/she, as in they and them, but plurals often make the rest of the construction harder and the meaning less clear. I used to recommend choosing one of the pronouns and the masculine gender in particular, where the intent is truly generic. However, I don't like that any more since some readers may interpret it as exclusionary. In other cases, you can introduce a distinct pronoun (he/she/they) for each character in your example and simply use that pronoun where necessary.
5. They (or them) when you mean a singular: as long as you can avoid ambiguity, but see the above. OK to use it in speech, though.
6. Etc. (me): except in casual writing—OK for emails, but not for papers. It indicates a carelessness of thought—a veritable verbal handwave. If there are more items that you need, state them; if not, let them be. You can convey an open-ended series by saying "A, B, and so on."
7. "And so on and so forth"—I don't see the point of going beyond "and so on." It seems the only reason people tack on the "and so forth" is to adhere to a cliched expression.
8. Though, at the beginning of a sentence: use although instead.
9. He or she for an artifact such as an agent: use it instead.
10. It for a person, even if used as an example agent. For instance, if you bring up Alice as an example agent, you should refer to her as "she" not as "it."
11. Issue (Lyn Dupre): usually a content-free word.
12. Actually, basically, really: often such adverbs are content free and suggest that the rest of what you said is not actual, basic, or real. Many uses of essentially are the same way.
13. Main or mainly, when used in a content-free manner, as it often is.
14. Kind of, sort of: meaningless.
15. Relatively, as in "this is problem is relatively easy." This usage is vague at best. Usually, it is because the writer has some comparison in mind, which he is not revealing to the reader.
16. Very (Mark Twain): usually content-free and frequently with a weakening, not a strengthening, effect.
17. At this point in time: need I say more?
18. Consistency, when you mean uniformity (me and possibly others): in our little world, consistency means logical consistency.
19. Different than: instead say "different from." Some weird grammar books will tell you "different from" is wrong. They are wrong.
20. Feel, as in "we feel that ..." Who cares about your feelings? "We believe" is better, but the best option is not to add such phrasing if you can live without it.
21. Believe, when talking about others as in "some researchers believe that the evidence is conclusive." How do you know what they believe? It is more defensible to say "some researchers claim that the evidence is conclusive" [and cite them, of course, to show that they make the claim you ascribe to them].
22. Lesser than: should be less than. It is OK to say "the lesser of the two evils."
23. The abbreviation cf or c.f. Classically, this indicates contrast, but enough people confuse it with e.g. that it is not a good idea to rely upon its correct meaning to make your point. Simply avoid it.
24. Have to, when you mean must. In informal writing, such as this document, it may be OK to say "have to," but not in formal papers.
25. Have and do, when you should be using more meaningful verbs. For example, you might replace "the agent does an action" by "the agent performs an action" and "we have a definition" by "we define" or "we postulate," and so on. A little bit of thought usually yields a verb that carries a more specific meaning.
26. Use a specific verb instead of a general one. For example, instead of "Bob performs an action" you can say "Bob acts"; instead of "Alice obtained a high rank" you can say "Alice ranked at the top" or "Alice ranked in the top quartile" (the latter further removes the ambiguity of a high rank).
27. Avoid stative verbs like "is" and "are," which often only imprecisely indicate a just-complete action. Instead, you might find it possible to use "become" or "arise" or "occur." In the same vein, you might consider replacing "holds" with "makes" (depending upon the rest of the sentence).
28. While, when you mean whereas. While is potentially ambiguous, because it also indicates an ongoing activity. Consider "While I order tea, he orders coffee." Is this is a contrast between tea and coffee or just a statement that the two ordering events occur concurrently?
29. As, when you mean whereas. Same point as above. There is another usage, which is fine: "Foo is bar, as remarked above" or "As remarked above, foo is bar." In other words, it is describing something outside of the scope of the document that can cause problems.
30. Little, when you mean a little (or vice versa). "There is little value in this approach" is almost the opposite of "There is a little value in this approach."
31. Few, when you mean a few (or vice versa). "Few people attended" means just a few (if any) did, whereas "A few people attended" means that at least a few did. This distinction becomes sharper when you modify this further. "Quite a few" means a lot.
32. Called as or considered as, as in X is called as Y. Just called or considered is what you need.
33. Prepositions when you can avoid them. Words such as "for" and "or" sometimes allow multiple interpretations. Where there is a clear(er) relationship, you ought to use a more precise word to capture that relationship.
34. Gerunds in compound nouns. You should not say "our motivation is to simplify service locating (or accessing or executing)"—either go for "service location" or "locating services." Acceptable exceptions would be phrases such as "resource sharing," I suspect, either because they are extremely routine or because there is no other noun corresponding to "sharing."
35. Match singulars and plurals— this concept is a form of agreement in linguistics (there are similar constraints for tense).
    • For an individual agent, use "it"; for many, use "they." It is really quite simple, yet many people seem to get this wrong.
    • In English, verbs convey the count, often through the use of an "s" suffix. However, an "s" marks a verb as a singular, whereas it marks a noun as a plural. "This module ensures" is OK as is "These modules ensure" but the remaining arrangements of "s" are wrong.
    • "Each" and "every" take singulars so "each agents" is wrong
36. Singular for plural. We don't say "commitments is a means ..." Instead we say "commitments are a means ..." Notice that the "are" goes with the plural commitments. Although several, the commitments are still "a" single means to an end here. However, "the essential elements of a multiagent system are the commitments among its member agents" is more natural than "the essential element of a multiagent system is the commitments among its member agents."
37. Improvise, when you mean improve.
38. Simple as, when you mean as simple as. In general, you need two copies of "as" to make such comparisons work.
39. Some, as in "some agent sends a message"—should be "an agent sends a message." Some conveys a vagueness. It can be combined with a weak verb to produce phrases like "some experiment will be done."
40. There is or there has been, at the start of a sentence or phrase. For example, "there has been some work done in security" is virtually meaningless. You can improve it to "some work has been done in security," which is still not highly surprising, to "relevant work in security includes the following," which leads us to what we care about.
41. Fitted, if you mean fit for or suitable. Fitted is what you are by a tailor. Agents aren't fitted for Web services, but are suitable for them. Agents can be fitted for Web services if you install something special in them.
42. Also: a fine word, but shouldn't show up in every sentence.
43. On, when it can be replaced by upon. Not doing so would not be an error, but I prefer to have fewer copies of "on" since it is a common word. Using "upon" makes the given phrase stronger.
44. Lots of when you mean several. It doesn't fly in formal writing. And don't say "several" if you mean "one" or "two."
45. Get, as in "to get a high rank." It is better to use obtain.
46. Firstly, secondly, and so on where you mean first, second, and so on. Often, you can count your points as One, two, and so on. I tend to use this form more often than first and second, but firstly and secondly is generally excessive.
47. Furthermore: further works just as well.
48. Faculty, as in "he is a faculty." A faculty is the entire body of teachers at a college or university. Some universities name each of their colleges "Faculty of X."
49. Monotonous when you mean monotonic. Monotonous means boring whereas monotonicity is a mathematical property.
50. The following words are singular, although they have the misfortune of ending in an s: semantics, pragmatics, pathos, ethos, characteristics (a curve, as of a transistor), kudos, bona fides, ... Educated people should get these right. On some occasions, you might make a concession to your readers and use the form they would understand, but I personally don't like doing so, idealistic academic that I am.
51. Implies or entails, casually. These are technical words, especially for computer scientists. You may even have defined one of them in your document. So use with care.
52. A code, when you mean a program. It is OK to use "code" as in source code, but then it is a mass noun. You don't have three "codes"—you just have code, but you may have three programs. Professor Jay Misra of UT Austin makes the point that your programs should be clear and understandable: if they really are like code, you aren't doing a good job.
53. Allow, unless you mean allow. Almost all cases of "allow" that I see don't mean allow. They merely reflect a paucity of thought. They are along the lines of "a coffee maker allows me to make coffee." It doesn't although it may enable or facilitate making coffee. Choose the right word.
54. Deduce, unless you mean deduce. Arthur Conan Doyle's use of the term as in "Sherlock Holmes deduced something" was incorrect. Inference is a better and more general term. Most inferences are not deductions. In natural settings, almost no inference is a deduction.
55. System, unless you identify which specific system you are referring to. Just about anything might be a system, and in computer science especially there are several reasonable possibilities.
56. Names of standards, such as SQL, XML, RDF, RDF Schema, except to refer to the standard or to use as a modifier. For example, you might refer to "an SQL query", "an XML snippet", "an RDF assertion", "an RDF Schema vocabulary", but you should not say "an SQL", "an XML", "an RDF", "an RDF Schema".
57. Alternate versus alternative—both are fine words: the former indicates changing back and forth as in left and right and odd and even; the latter indicates choice.
58. Keep versus put.

Words and Phrases to Treat Uniformly

Use the following words and abbreviations uniformly within any document, because there seems to be no hard and fast rule for them:

1. Life cycle or lifecycle.
2. Health care or healthcare.