320 likes | 364 Views
Writing for CS and CE Research Paper. CSCE 481 FALL 2015. Acknowledgment – Prof. John Keyser & Aakash Tyagi. Specific Writing. We will discuss some specifics for writing research papers . This is commonly done in graduate school. Material developed for graduate students
E N D
Writingfor CS and CEResearch Paper CSCE 481 FALL 2015 Acknowledgment – Prof. John Keyser & Aakash Tyagi
Specific Writing • We will discuss some specifics for writing research papers. • This is commonly done in graduate school. • Material developed for graduate students • But, many principles carry forward to other writing • Other topics: • Proposals • Technical documents • Executive summaries
Sections of a Research Paper • Title • Abstract • Introduction • Previous Work • Possibly including background information • Main Work (ideas/theory/exposition) • Possibly in several sections • Implementation • If needed • Results • Possibly combined into main work section • Conclusion • With future work • Acknowledgements • References • Appendices
Title • Don’t underestimate title importance • Memorable titles can help people remember the paper • The title will be used for searching, later • Remove unnecessary words • Watch for misleading words
Introduction • Motivation and Summary • By the end of the introduction, someone should be able to tell someone else what you did, and why. • But probably not give any details about how • Keep the introduction short, relative to the rest of the paper.
Motivation • Early on in the paper, you must make the case for why you are doing this • This should not be too long • If you have to spend too long to say why someone should read the paper, then there’s probably not a good reason • The motivation is not why you are writing the paper, it’s just there to get people to read it • Sometimes this is more important than other times – sometimes motivation is obvious
Summarizing Main Results • You want to make it clear what the main results of your paper are. • Don’t “hide” them or make them a “surprise” at the end • Remember, most people will not read your full paper – you still want them to know the main results • Should always be in the abstract • Should be in the introduction of the paper • Main Results, Contributions, Thesis Statement • Can be in the conclusion
A “Main Results” Section • Could be a subsection, a paragraph, a bulleted list, or a sentence • Should be easy to find/locate • Should make clear what is the new, unique contribution of this work • It is not a summary of everything you’ve done, or even a summary of the paper • Just list the key point(s) that are new to your work.
The Thesis Statment • A short statement that summarizes what the focus of the paper is • Can help to focus your writing, presentation, and research • The goal of the paper is to show why the thesis statement is important and true (or false…)
Previous Work Section • Provide references to relevant material • What are the key papers that someone should read to understand this? • What are the most relevant related papers/alternatives? • Demonstrate that you are familiar with the main research in the area • Ensure you cite all the relevant work • Especially the papers of those who will read yours… • Can’t cite everything; cite the most important things • Usually, citations to textbooks aren’t needed • Unless that textbook provides a unique derivation, a particular summary, etc.
Previous Work Section • If necessary provide background summary of prior work • For example, if you are building on your own prior work • Make sure that prior work is separated from new work • You want to clearly delineate what is new vs. what is old. • When giving citations to previous work, it is good to show how your work fits in with that prior work.
The New Work • This is the main, core part of your paper • It should be the part that you are most confident in, and have the most to say about • It is important that you are clear and accurate.
Things to Look Out For • You are not just presenting a list of what you did. • Every piece of research has lots of “infrastructure” work that goes on behind it – you don’t need to go into this, unless it is critical • You don’t need to discuss “dead end paths” that you pursued • One exception is if it is very likely someone else would follow that dead end path • You research is evaluated on results, not process.
Things to Look Out For • You want to develop your material clearly • Usually, someone will read this section in order • Don’t pull ideas/material from nowhere • Make sure that information is presented in a logical order • Think of it as telling a (technical) story: • Keep the story moving • Don’t refer to things that the reader has no knowledge of • Make sure the reader understands what has happened!
Things to Look Out For • Avoid tangential topics • Make the section about the main results, not the interesting “side” items • Use appendices if necessary • Make sure there is a clear overview • Avoid going directly into details if the person doesn’t have the overall picture • Often, overview sections or figures are helpful
Results • You want to demonstrate all of the core ideas that you discussed in practice • If you discussed something, show the results • Idea is to show that what you presented works, and give some sense of how well it works • Pick good test cases, that cover a range of situations • Ones that allow comparison • Ones that allow evaluation of parts of your technique • Ones that simulate “real world” cases • You need to provide comparisons to other work, whenever possible • This lets people evaluate your work
Conclusion • Now that we have seen the work in the paper, what can we conclude? • What has been the “contribution” of this work? • What insights does this work offer? • What does this now allow us to do? • Conclusion should not be just a summary of what was in the paper – that is obvious.
Future Work • Usually part of the conclusion • Not always included, but a good idea if possible • People want to know that the paper is not a “dead end” • What more could be done? If I like this area, what could I work on next? • Is this likely to stimulate future work? • Can be a “defense” against reviewers.
Future Work • Avoid using “throwaway” future work • In computer science, you can always say you want to improve performance, port to a new system, or integrate with something else. • Better to have one or two solid areas for future work than 10 that aren’t developed. • Don’t just state areas, give some indication of the challenges/opportunities • Why will that be worthwhile? • What are some obstacles that will be faced in that extension?
Audience • Make sure you are writing to the appropriate audience • Usually, this is to other researchers in the field • Not to novices – they will know the basics of the field • Not necessarily to just the foremost experts in the area – they will not be familiar with every bit of prior work • Not to experts in all areas – they may not be familiar with simpler concepts from other fields • Some papers (e.g. literature reviews) are for more general, less expert, audiences
Audience • Give them the background they need to understand the paper • Particularly if you rely on another technique; don’t make them read other papers before they can read yours • Not always possible – sometimes there is too much to do • Notation might not be standardized • Explain the notation as needed • The concepts might already be known
Overstating/Understating • Do not oversell your work • Do not promise more than you deliver • Do not try to make your work have more impact than it reasonably does • You probably have a higher opinion of your work than others do or ever will. • Readers are annoyed if they spend their time reading your article, only to find it didn’t do what was promised.
Overstating/Understating • Do not undersell your work • Don’t put in so many disclaimers that you discourage someone from reading/following it • Point out problems, especially key ones, but: • Your goal is not to point out every conceivable flaw • If necessary, point out why problems might not be so bad • You are writing the paper because you have something new to present, that others should find valuable.
Overcoming Objections • Those reading the paper will often have questions/objections. • You want to answer/address these in the paper • This is key to getting the paper accepted through review, but also for getting the paper accepted after publication
Overcoming Objections • Think: “If I were a reviewer, what would I have questions about?” • Find a way to address those directly • If they are technical concerns and you have not addressed them in the work, show that you’ve thought about them • What examples should be included? • What tests should be provided?
Figures and Captions • People will usually look at figures before they read the text • You want the figures to stand on their own as much as possible • Be sure that your captions clearly describe what is in the figure. Do not rely on the text to describe the figure.
Comparisons to Prior Work • Always a tricky proposition • Your goal in the paper is to show how good your work is. You have spent a great deal of time on your own approach. • You must be fair to prior work, but you probably can’t devote as much effort to replicating it. • If standardized comparisons can be made, use them • If you implement another method for comparison, be sure to do your best with it • If not, be sure to clearly state what you did not do, and why.
Comparisons to Prior Work • It is not OK to just present your material and assume it should be accepted • That does not show any new contribution over the state of the art • Exception: if it is truly the first time someone has accomplished something • If you cannot provide comparisons, at least provide concise, clear arguments that evaluate your method vs. other methods.
Feedback • If possible, get someone else to read your work • They should be willing to give direct, honest feedback • Take their evaluations to heart • When reviewers reply with objections, don’t blame the reviewer • If the reviewer didn’t understand it, it’s probably your fault • Make sure that you address their concerns • Sometimes it is only a style/writing issue! • Sometimes they have found more fundamental flaws • Even these can sometimes be addressed by writing differently. • There are (very rare) exceptions where reviewers are way off • Always be polite and respectful in your responses, anyway