Advice on understanding related work

Contents

How to do research

How to write a paper

This is actually about how to write a professional looking paper, and not so much about how to write well. I may add more of the latter as time goes on, but I'm still learning how to teach it. Why is this stuff important? Caveats: Most of you are still learning to write technical papers. I know that. This section is mostly not about that (although it's something that will need continued work). Instead, it's about all the things that you could do even if you knew nothing about writing, to make the process smoother and more successful. In the future, I will likely return drafts unread that do not address the issues raised below.

The above list may seem daunting and unecessary, after all the work that went in to the actual content of a paper. You may even think, after having worked with me on a paper, that I am too much of a perfectionist. It is not unecessary, although it does take time. I do have high standards. However, refer back to my comment on professionalism. Papers submitted to conferences should be ready to publish. This is crucial to understand. The impression you make on reviewers depends in large part on meeting this bar.

In addition to the above issues, here are some writing tips. The following things should be done from draft 1:

The following is a tip that you will learn to incorporate over time, but should not be a focus in your early drafts. This tip is an easy fix later on, and likely to cause writer's block if applied to early:

Say what you're going to say, say it, and then say that you said it. Another piece of advice is that every unit from the entire paper down to each paragraph should have that structure. I should be able to look at just the abstract and get a sense of the whole paper. I should be able to look at the first paragraph of each section to get a sense of what the section is about. I should be able to look at the first sentence of each paragraph and learn what the paragraph is about.

Another tip I just found myself writing down in response to a paper I just read: My rule of thumb for related work is only say as much as you have to. That means that you only describe something in detail if (a) understanding it is key to understanding your paper or (b) a reviewer might question how much your work differs from the work you cited. Otherwise, you need to practice the skill of summarizing why it's relevant without explaining the whole system. Careful not to simply say *that* it's interesting though.

Some other advice

How and when to proof read a paper

A paper should be proof read fairly often -- I typically do it at least 6 hours after I last edited it, and always before handing it to co-authors. The following is the steps I go through when doing a thorough proof. First, I make a list of things relevant to that particular document. For example, here's a sort of generic list that I often copy over. After I have that list, I try to keep it in mind when I write, but I also use it to do a final pass over a document before submitting it.

main paper

-------tougher ones----------

references

Finding Papers

How do you find important papers for a topic you're interested in? First, you need to figure out the right "keywords". You do this by searching for papers with different keywords until you find at least one paper that's relevant. That's all you need.

Next, you need to figure out how to find other similar papers. Look at the keywords in the paper, write down the ones that relate to your problem. You can use them to search for other papers. Look for the author's home page. Figure out who the authors are. Who did which part of the work? Who is the advisor, who are students or collaborators? What are the research areas of these people? Likely, one or more of them will have other relevant publications.

Finally, trace the tree. Any decent papers will reference other relevant work. You can find this by looking at the reference section at the end of the paper. Also, many papers (especially if they are older) will be referenced by other work. You can find referring papers using what's called a citation index. Comprehensive citation indexes can be found in libraries (there's a link to an online one below).

You also want to find the actual text of papers, rather than just the reference and abstract info. These days, in computer science, you rarely need to go to the library. If a search at one of the sites below doesn't turn the paper up, try looking for the home pages of authors. They usually have an electronic copy available (or might be willing to email one to you if you ask).

Finding Other Related Work

Although finding papers is a critical starting place, it is important to make use of other sources of information including market research and studies of your own. Below are some links to interesting data sets.

Reading and Reviewing Papers

One important first step here is to determine quickly whether and how deeply you should read a paper. Skimming is an important skill. Start with the title, then the abstract. Then jump to the conclusions. If you're still interested, read the introduction, and if you need more details, only then do you actually read the paper.

If you get to this stage, always write up your thoughts in a summary, and keep a copy of the paper on file. The links below really do a good job of telling you how to write such a summary. Just remember, as you read papers in your field, you are trying to build up not only an understanding of each paper, but also an understanding of the landscape in which they sit. Who are the authors? Have they done groundbreaking work in this area, contributed incremental next steps, or are they perhaps just starting out (maybe they're still students). What about this paper? Is it groundbreaking now? Was it groundbreaking when it was published? Why would someone want to read it?

One last point: if you are doing research, you are looking for the gaps in the field. Every paper has a future work section that suggests possible gaps. Have they been filled in? Do any of them spark your imagination?. You can also think about a paper as suggesting a lense through which to view things. If you read a paper about cscw, and are working on an ambient display project, think about how the issues raised in the cscw paper might apply to ambient displays. The reverse applies too of course. A lense like this can also help to identify gaps, things that need to be explored and discussed.

Presenting a paper

You will have to present many papers over your years as a researcher, both your own work and that of others. For many, this does not come easily, but will improve with practice, and can be alleviated with adequate preparation.

Understanding related work in your field

Do you know about other research groups working in your area and where you and your work fits in? Some of this can be found by reading related work, some of it through conferences. Variously in the past, I've been part of groups that keep a web page of relevant conferences/journals (this helps suggest where to publish/look), make web pages of related work, and write literature reviews of related work that try to address these sorts of things. In the end you build up a picture in your head that is sort of a graph of people and projects and who spawned/taught what and moved where. This is also what you'll be doing for the field of HCI in general as you prepare for your qualifiers.

This kind of process is an important form of networking, background research for a PhD thesis, and generally benefits research. Once you've done this, you know who to try to meet at conferences, where to look for possible new work, and when you want to, say, sponsor a workshop on a topic, or find a summer internship, you've got the right contacts/people to invite. And it helps you to see where your own work fits in and how it is different. Always think about papers from this perspective as you read them.

Resources in HCI

You should be aware of resources in your are. They vary from organized information about references such as HCI Bib and other links mentioned above on this page to discussion and mailing lists. In particular, you should consider joining CHI-STUDENTS, CHI-ANNOUNCEMENTS, and ACM in order to stay in touch with your larger research community.

The Job Search


Advice for once you're graduated and employed

jmankoff at cs dot cmu dot edu