This web page is my attempt to collect together a few useful pieces of advice for writers, especially writers of research papers. All of these tips are focused on improving individual sentences. If you follow these tips, you should hopefully end up with a clearer and easier to read paper, and you will almost certainly reduce the length of your paper as well (always good for conferences with length limits).
For more detailed explanations as well as information on how to improve organization and cohesion between sentences and paragraphs, I highly recommend the book Style: Toward clarity and grace by Joseph M. Williams. (There is also a more recent version called Style: Lessons in Clarity and Grace; I'm not sure how much it differs.)
First, take a look at these tips on writing clear, concise sentences from the University of Wisconsin Madison writing center.
Below are some further examples of phrases edited according to their suggestions. These closely follow actual examples from text I have written (and then edited) or read in papers. I've divided them into categories similar to the ones used by the Madison center so you can read their explanations, but really many of these guidelines follow from the very simple idea that you should use strong verbs as much as possible to describe what's going on. A strong verb is normally specific (rather than vague), active (rather than passive), and a single word (rather than a verb-particle construction or a whole phrase). It is also not a noun.
X is the instantiation of Y → X instantiates Y
X is an indication that Y → X indicates that Y
one observation is that → we observe that
provides an explanation for → can explain
has an impact on → affects
provides a justification for → justifies
the use of X can → using X can
we make use of X → we use X
the application of these techniques can → applying these techniques can
it is our hope that → we hope that
have more general applicability → are more generally applicable OR apply more generally [In this case the adjectival form 'applicable' is arguably better than the verb 'apply'; there are always exceptions to any rule]
Three experiments were run → we ran three experiments
several variations were evaluated → the authors evaluated several variations
our model is described in section 3 → we describe our model in section 3 or section 3 describes our model.
[You may have been told at some point to avoid "I" and "we" in formal writing, but this depersonalized style really isn't required practice in computer science or psychology. Using passive voice just in order to avoid first-person pronouns often makes sentences convoluted and in some cases actually introduces ambiguity about who performed the action. However, if you really want to avoid first-person, notice in the second two examples above it is often possible to avoid the first-person while still using an active sentence. As a final note, there are also cases where passive sentences are completely clear and appropriate; as with most if not all the advice on this page, there are no absolutes.]
Verb-particle constructions are multi-part verbs like take off, put up with, throw up, look over where the meaning is determined by both parts. Try not to insert too much material in between the verb and particle, or use a more descriptive single-word verb that doesn't depend on a particle for its meaning:
we turned the left-branching structures into right-branching structures → we converted the left-branching structures into right-branching structures
we threw all examples of length 2 out of the data set → we threw out all examples of length 2 from the data set or we removed all examples of length 2 from the data set
Examples of weak verbs include go, have, make, take, put, and many others with little specificity. These verbs usually require particles, longer phrases, and/or nominalized verbs (see 'put the action in the verb' above) to make them specific.
We chose to (decided to, elected to) implement → We implemented
Our results provide evidence that → Our results suggest that
works by iterating → iterates
comparable results on f-score→ comparable f-score
goes into more detail about → details
provides support for → supports
Note the last two are also examples of 'put action in the verb' though the noun forms are not morphologically complex nominalizations, so not quite as egregious as many of the examples in that section.
Unnecessary adverbs and adjectives and wordy hedging phrases are often used to qualify statements in scientific papers. Practice eliminating them.
can be seen/viewed/considered/thought of as → is
is essentially/basically → is [unless it isn't, in which case explain what's different instead]
It is worth noting that X → Note that X or X
We wish to emphasize that X → We emphasize that X or X
Based on analysis of our results, we argue → We argue
a particular advantage → an advantage
an especially interesting result → an interesting result [and: is it really interesting? Or can it just be 'one finding'?]
In addition to 'especially' and 'interesting', watch out for other intensifiers and subjective adjectives and adverbs, such as:
Suprising(ly), interesting(ly), remarkabl(y), unusual(ly)
Very, extremely, exceedingly, vastly, completely, relatively, especially
When using comparative phrases, make sure it's clear (a) what is being compared to what, and (b) on what dimension. Common problems include:
similar to: This phrase is often used by students who want to avoid precision and aren't really sure what the similarities are (as well as by students who know perfectly well but are just being lazy). If you say X is similar to Y, you'd better make sure it's clear what the similarities (and differences!) are.
better than: Make sure you've stated what makes something "better". More efficient? Higher accuracy (on what measure)? Simpler to implement? Consider replacing "better" with one of these more specific phrases.
This method/result/etc is more X.: Even when writers are specific about X (efficient, accurate, etc.), it's all too common to leave out the other half of the comparison. More X than what? Sometimes the comparitor is clear from context, but often it isn't nearly as clear as you think. Think carefully before using comparisons without 'than'.
Too often, "this" by itself is used to refer to some underspecified idea or issue raised in the preceding context. Even when there are concrete concepts under discussion, "this" can be ambiguous: which concept is the one you are talking about? It's often less clear than you think. To avoid these problems, follow "this" with a noun, or replace with a noun phrase. If you can't figure out what noun is appropriate, you almost certainly need to do more serious editing.
To avoid this → To avoid this problem
This shows that → These results show that
This means that → This fact means that
This gives us → This model/parameterization/approach gives us
This is challenging because → This task/problem is challenging because
Often a single sentence has multiple possibilities for improvements. Here are a couple of examples.
Here's another example where each step makes multiple edits to improve on the previous version.