Writing Style

1. Writing Style Specific guidelines

2. Titles and headings Direct statements Emphasis Sexist language abbreviation Sentence structure Repetition and parallelism Ambiguity Qualifiers Misused words Redundancy and wordiness Outline

3. Titles and headings (I) • Rule 1: be concise yet informative • An investigation of the Effectiveness of Extensions to Standard Ranking Techniques for Larger Text Collection. • Extensions to Standard Ranking Techniques for Larger Text Collection. • Rule 2: be specific • Huffman Coding for Databases. • far too general. • Limited-Memory Huffman Coding for Databases of Textual and Numeric Data. • Specific, superior to the previous one, but awkward. • Limited-Memory Huffman Coding for Databases.

4. Titles and headings (II) • Rule 3: not necessarily to be complete sentence • Duplication of Data Leads to Reduction in Network Traffic. • Duplicating Data to Reduce Network Traffic. • Rule 4: it is not usual to put a stop at the end of a heading • Neutral Nets for Image Classification. • Neutral Nets for Image Classification • Rule 5: section headings should reflect the logical structure • E.g., 3. Lists and Trees and 3.1. Lists3.2 Trees.

5. Direct statements (I) • Avoid excessive use of indirect statement (also known as passive voice). The direct style (or active voice) is often easier to read. • The following theorem can now be proved. • We can now prove the following theorem. Comment: • If passive voice is necessary, use it. • Use of “we” is valuable. • Consider “we show…” versus “in this paper it is shown…” • “this paper shows…” and “this section argues…” are often used by our Chinese people. Strictly, they are not a very good phrases, better be replaced with “in this paper we show…” and “we argue in this section…”, respectively.

6. Direct statements (II) • Even in the paper of only one author, the use of “we” is acceptable. • In some case the use of “we” is wrong. • When we conducted the experiment it showed that our conjecture was correct. • The experiment showed that our conjecture was correct.

7. Emphasis (I) • Emphasis (or stress) on some words can be implicit or explicit. • Implicit emphasis is implied by the structure of the sentence • Compare “it is not a good idea” with “it is certainly not a good idea”. • Reorganizing a sentence can change the emphasis • The algorithm is appropriate because each item is written once and read often. • The algorithm is appropriate, because each item is only written once but is read often.

8. Emphasis (II) • Inappropriate stress can lead to ambiguity • Additional memory can lead to faster response, but user surveys have indicated that it is not required. The stress on ``additional memory’’, incorrectly implies that users has commented on memory rather that response. Since the sentence is about ``response’’, that is where the stress should be: • Faster response is possible with additional memory, but user surveys have indicated that it is not required.. • Explicit emphasis is provided with italics, CAPITALS. • DON’T use capitals for emphasis. • Use italics only to stress a word or a phrase. • The first time a key word is used, consider placing it in italics.

9. Sexist language • Some readers find use of “he” or “his” for generic case offensive. • A user may be disconnected when he makes a mistake. • Sexist usage is easy to avoid. • A user may be disconnected when they make a mistake. • Such use of “they” as a singular pronoun is acceptable but jarring (不和谐). • A user who makes a mistake may be disconnected. • Note: don’t use constructs such as “he or she” or “s/he”.

10. Some common abbreviations in science writing: no.=number i.e.=that is e.g.=for example c.f.=compare with (perhaps more accurately “in contrast to”) w.r.t.=with respect to W.L.O.G=without loss of generality iff=if and only if etc.=and so on Fig.=Figure Alg.=algorithm These save a little space, but slow readers down. “don’t”, “isn’t”, ect. are not suitable for science writing. Abbreviations (I)

11. A slash (/ also known as a virgule or solidus) is often used for abbreviation, as in Save time and/or space Used for list/tree processing confusion: /=either but not both? (in the usual English sense?) /=either or both? (in the usual computing sense?) /=and? /=also? If you want to be clear, don’t use slashes. Abbreviations (II)

12. Abbreviation by acronym (取首字母缩写) CPU for “central processing unit” OS for “operating system” DBA for “data base administrator” … Abbreviations are usually terminated by a stop, but it is unusual to put stops in acronyms Thus “CPU” is correct, “C.P.U.” is acceptable but pedantic（学究）, and “CPU.” is incorrect. Plurals of acronym: just append an s Write CPUs rather than CPU’s, OSs rather than OSes Abbreviations (III)

13. Sentence should have simple structure, usually no more than a line or two. Don’t say too much all at once. Sentence structure (I) • When the kernel process takes over, that is, when in the default state, the time that is required for the kernel to deliver a message from a sending application process to another application process and to recompute the importance levels of these two application processes to determine which one has the higher priority is assumed to be randomly distributed . • When the kernel process takes over, that is, when in the default state,the time that is required forthe kernel to deliver a message from a sending application process to another application process and to recompute the importance levels of these two application processes to determine which one has the higher priorityis assumed to be randomly distributed . should be explained elsewhere Too long (34 words). • When the kernel process takes over, one of its activities is to deliver a message from a sending application process to another application process, and to then recompute the importance levels of these two application processes to determine which has the higher priority. The time required for this activity is assumed to be randomly distributed . Only two words

14. Sentence structure (II) Watch out “if” expression • If the machine is highly loaded then speed is acceptable whenever the data is on local disks. • If the machine is highly loaded then speed is acceptable whenever the data is on local disks. The consequent The two conditions are separated by the consequent. • If the machine is highly loaded and data is on local disks then speed is acceptable. • Speed is acceptable when the machine is highly loaded and data is on local disks. • If data is on local disks then speed is acceptable, even if the machine is highly loaded.

15. Beware of misplaced modifiers Sentence structure (III) • We collated the responses from the user, which were usually short, into the following table. Which=the user, response, or the collation? • The user’s responses, most of which were short, were collated into the following table.

16. Double negatives are difficult to parse and are often ambiguous Sentence structure (IV) • There do not seem to be any reasons not to adopt the new approach. The impression here is of condemnation—we don’t like the new approach but we’re not sure why. • The new approach is at least as good as the old and should be adopted.

17. Complementary concepts should be explained as parallels, or the reader will have difficulty seeing how the concepts relate Repetition & parallelism (I) • In SIMD, the same instructions are applied simultaneously to multiple data sets, whereas in MIMD different data sets are processed with different instructions. SIMD and MIMD are complementary concepts. • In SIMD, multiple data sets are processed simultaneously by the same instructions, whereas in MIMD different data sets are processed with different instructions.

18. Repetition & parallelism (II) Parallel can be based on antonyms • Access is fast, but at the expense of slow update. • Access is fast but update is slow. Lack of parallel structure can result in ambiguity • The performance gains are the result of tuning the low-level code used for data access and improved interface design. • The performance gains are the result of tuning the low-level code used for data access and of improved interface design. It is kinder to the reader to move the longer clauses to the end • The performance gains are the result of improved interface design and of tuning the low-level code used for data access.

19. Ambiguity (I) Check carefully for ambiguity. • There is a new version of the operating system, so when using the “fetch” utility, the error messages can be ignored. • There is a new version of the operating system, so the “fetch” utility’s error message can be ignored. • In addition to lists we have also tried trees, they are superior because they are slow in some circumstances but have lower cost. • In addition to lists we have also tried trees. Lists are superior because, although slow in some circumstances, they have lower cost.

20. Ambiguity (II) Plurals may cause ambiguity. • Packets that contain an error are automatically corrected • Packets that contain errors are automatically corrected The first version implies that packets with a particular error are corrected, the second impliesthat packets with multiple error are corrected. Both of these interpretations are wrong. • A packet that contains an error is automatically corrected.

21. Qualifiers should not be piled on top of each other. Qualifiers (I) • It is perhaps possible that the algorithm might fail on unusual input. Words such “might”, “may”, “perhaps”, “possibly”, “likely”, “likelihood”, and “could” can be used once in a sentence, but not more. • The algorithm might fail on unusual input. • We are planning to consider possible options for extending our results. • We are considering how to extend our results. • Merten’s algorithm is not dissimilar to ours. Such sentence tells the reader little. Double negatives are a form of qualifier commonly used to express uncertainty.

22. Qualifiers should not be piled on top of each other. Qualifiers (II) • There is very little advantage to …. • There is little advantage to …. • The standard method is simply too slow. • The standard method is too slow.

23. Misused words (I) Which & that • There is one method which is acceptable. • There is one method that is acceptable. • There are three options, of which only one is tractable. • It is true the result is hard to generalize. • It is true that the result is hard to generalize.

24. Misused words (II) May & can: Use “may” to indicate choice, and “can” to indicate capability. • Users can access this facility, but may not wish to do so. Less & fewer: Use “less” for continuous quantities, and fewer for discrete quantities. • It used less space. • There were fewer errors. Alternate & alternative: The word “alternate” means other or switch between, whereas an “alternative” is something that can be chosen. If there is but one alternative, there is no choice. “alternative” and “choice” are not synonyms.

25. Misspelt words and misused words Misused words (III)

26. Redundancy and wordiness (III) The table lists common redundant or wordy expressions and possible substittutes for them

27. Diversity—in organization, structure, length of sentences and paragraphs, and choice of words—is a useful device for keeping the reader’s attention. Variation • The system of rational numbers is incomplete. This was discovered 2000 years ago bty the Greeeks. The problem arises with squares whose sides are of unit length. The length of the diagonals of these squares is irrational. This discovery was a serious blow to the Greek mathematicians. • The Greeks discovered 2000 years ago that the system of rational numbers is in complete. The problem is that some quantities, such as the length of the diagonal of a square with unit sides, are irrational. This discovery was a serious blow to the Greek mathematicians. In the second version, the final statement is more effective although is hasn’t changed.