American editors use Strunk and White’s little book of rules to check the elements of style. Read it, read it again, and again. Late editions do not improve over early editions.
If you need a verbose description that addresses technical issues too, look for "Bugs in Writing."
I will use the following dimensions to evaluate your descriptive memos:
organization : does the essay come with an overview paragraph that states the overall thesis and spells out what the rest of the essay will cover?
paragraphs : are the paragraphs organized properly? A paragraph, like a section or an entire essay, is a unit of writing. As such, it must come with a thesis and focused, single-minded support for the thesis. This support can come in various forms and shapes: logical arguments, general explanations, and examples.
Paragraphs should not be dialectic, i.e., they should not cover both thesis, antithesis, and synthesis. If an antithesis is to be presented, the paragraph should contain a clue word in one of its last sentences, e.g., "however", and a sentence that transitions to the next paragraph.
conclusions : is the conclusion supported by the essay? or does it inject new ideas? Don’t write it in a high-school-ish way.
flow: do sentences and paragraphs connect to each other? I will focus on sentence-level flow. One trick is to use a word/phrase from near the end of sentence n at the beginning of sentence n+1 and to move on to a new idea here.
writing : does the essay suffer from low-level problems here?
spelling: Here are some typically bad typos to watch out for: "it’s" (contraction of it and is) vs "its" (possessive) "their" vs "there". Also pay attention to capitalization.
choice of words: mismatch in numbers, "These hats is round." the use of non-descriptive/inappropriate nouns (wrong meaning), verbs (wrong connotation in context of surrounding noun-phrases), adjectives (highly subjective, colorful). It’s also common to find "it" or "this" without a clear reference.
sentences: Scientific writing calls for complete sentences.
tense: Scientific papers mostly describe eternal truths. When they do, they should use present tense, not future, not past. On occasion, a paper describes an experience, in which case past tense is appropriate, or future work (beyond the work explained in the paper), in which future tense is acceptable.
ideas/content : is a claim obviously flawed (check the web)? is a statement inconsistent with the rest of the claims? is terminology used in an obviously mistaken manner?
Below is a sample memo based on Matthias Felleisen’s experience with a paper by Plotkin which he encountered as a PhD student. He says it took him four months to understand the basics of the paper, and he then spent nearly ten years working out solutions to the implied problems that the paper poses.
Call-by-name, Call-by-value, and the Lambda Calculus
Gordon Plotkin’s paper  studies the relationship between programming languages and Church’s lambda calculus. It specifically addresses the scientific debate of how the call-by-value parameter-passing mechanism in most programming languages relates to the beta rule of the calculus. Concretely, the paper presents two variants of Landin’s abstract SECD machine for evaluating lambda terms—
one for evaluating according to call-by-name and one for call-by-value— and a variant of the lambda calculus for each of them. It then proves one set of theorems for each machine-calculus pairing: a Church-Rosser theorem, a Curry-Feys Standardization theorem, a theorem relating the standard reduction relation and the respective SECD, and a soundness theorem for the equational calculation system with respect to the observational equivalence relation generated by the SECD evaluator. The paper concludes with a section that relates the two calculation systems to each other via Fisher’s continuation-passing style transformation.
At this point, we fail to understand three aspects of these results. First, it remains unclear how the calculations in the by-name and by-value calculus differ for sample programs. Second, we are wondering how the equational calculation system relate to the standard reduction relation. Third, we wish to study Plotkin’s proof that the evaluation function based on the standard reduction function is the same as the one defined by Landin’s SECD abstract machine.
For the first two questions, we intend to build a call-by-name and a call-by-value calculus, respectively, with Redex. Then we will create a series of increasingly complex sample programs and study their reduction graphs, i.e., the graphs generated by the general reduction relation of the calculus. We expect that we can develop a strategy for identifying the standard reduction path(s) in these graphs. Finally, for the third question, we intend to break up Plotkin’s proof into smaller pieces, focusing on one abstract register from the SECD machine at a time. By transforming the machine one step at a time, the proof should become self-explanatory. We will initially focus on the proof for the call-by-value system; it should become obvious how to port this method to the call-by-name system.
1. Gordon Plotkin. Call-by-name, Call-by-value, and the Lambda Calculus. TCS 1974.
2. Michael Fisher. Lambda-Calculus Schema. Proceedings of ACM Conference on Proving Assertions, 1972.
Compare this memo with the proposed structure on the Project page.