Authors:
David Chiang (University of Notre Dame)
Kevin Knight (University of Southern California / Information Sciences Institute)
In our previous post, we ended with the advice to write for your readers, not yourself. This is, truly, easier said than done. Here are three concrete ways to put this into practice.
The (n + 1)st revision is better than the nth revision.
“I have rewritten – often several times – every word I have ever written. My pencils outlast their erasers.”
— Vladimir Nabokov
To write for your reader, you’ll need to read your paper — over and over again. Analyze each sentence, each paragraph, each section, from the reader’s point of view, not your own. Try to predict what the reader will ask, and answer it preemptively by fixing your paper. Every pass you make through your paper, you will spot mistakes in your writing – or, better yet, bugs in your ideas – and the end product will be better.
After you’ve revised your paper sufficiently, ask your friends to read it. They will spot dozens of problems that you missed. This may make you feel uncomfortable. But isn’t it better for your friends to see those problems than for the world to see them?
Backwards is better than forwards (sometimes).
“A talk is not a murder mystery.”
— Stuart Shieber
Good flow often puts things in the opposite order from what you might think. You might think you want to say things in logical order or chronological order, but that is not the order that is most useful to the reader. Murder mysteries reveal the identity of the killer at the end. Cookbooks do not reveal the end product at the end. (“Surprise! It’s a…foie gras tiramisu!”) Rather, they show you a picture of the end product at the beginning. Your papers should be more like cookbooks than murder mysteries.
- First state the main result in the introduction, and then tell how you reached that result in the rest of the paper.
- First say what you did, then say how you did it.
- First state a claim, then provide the justification for the claim.
- First give an example – people extrapolate surprisingly well from one example – then describe the general case. (But don’t skip the general case.)
- First give an intuitive explanation, then give the rigorous explanation. (But don’t skip the rigorous explanation.)
Shorter is better than longer.
“I have made this longer than usual because I have not had time to make it shorter.”
— Blaise Pascal
In Newfoundland, there is a stretch of highway that is home to many moose and is also very foggy, resulting in many car accidents. In the late 1980s, the authorities erected life-sized, reflective moose-shaped signs to warn drivers of this danger. The moose signs did reduce the number of accidents, but drivers began to slow down to look at the unusual signs or stop to take pictures with them, causing more accidents. The enacted solution was to place new signs a half-mile earlier that read: CAUTION: MOOSE SIGNS AHEAD.*
When you see a problem in your paper, you may be tempted to fix it by adding an explanation to it (“moose signs ahead”). But this may only prolong the problem (and push you further over the 8-page limit). Instead, follow your reader’s thought process as they try to navigate your writing, find the problem text, and replace it with something that serves your reader better, keeping in mind that sometimes the best way to make something less confusing is simply to delete it.
* “Moose Signs.” In Robert Finch, The Iambics of Newfoundland: Notes from an Unknown Shore, pages 114–116.
NAACL HLT 2018 Chairs Blog 