Tuesday, November 12, 2013

A Formula for Academic Papers: Title and Abstract

There are many ways that people will use the papers you write. Only a few will pick up your paper and read it straight through, from start to finish. More often, your readers will approach the paper with a focus on the specific component that will aide their research. Somebody running a similar study to yours, for example, may want to borrow your methodology or build on a particular finding you present. A person doing a literature search may want to place your paper in the context of other research or use it as a source of additional papers to read. Even those who do end up reading your paper straight through will probably start by scanning the conclusion, figures, and references to decide if the paper is worth the effort. In each of these cases, however, your reader will also read the title and abstract to develop the necessary context. You want to be sure that these form a clear and complete view your work.

Titles in HCI and IR sometimes consist of two components of the form Short Phrase: Followed by a Longer Description of the Paper Content. Nineteen of the last 45 papers I have published (or 42%) follow this format (e.g., Slow Search: Information Retrieval without Time Constraints or #TwitterSearch: A Comparison of Microblog Search and Web Search). This structure is so common as to seem somewhat trite, and people are quick to make fun of it. But I believe that it serves a useful purpose.

The initial short phrase provides those who are familiar with a piece of work an easy way to refer to it, and I have found that I miss it when absent. For example, my colleagues and I worked on a project that we refer to as “Tail Answers,” but the corresponding paper (Direct Answers for Search Queries in the Long Tail) does not include that term in the title. This makes finding the paper harder for interested parties (including myself, as familiar as I am with the work).

However, a short phrase alone is not sufficient for a paper title. The title also needs to include a good preview of what the paper will cover in the form of a longer description. I used to try to pick up space in my papers by cutting a two-line title down to one line, but I have since become convinced that it is much better to cut a paragraph of content from the body of the paper than a few words from the title. The first exposure may people will have to the paper will be through the title alone, perhaps in a conference program or in the reference list of another paper. The title needs to convince the person who encounters it to learn more about the paper without any additional information. For this reason it should provide as much context as possible (e.g., “Personalized Web Search” is better than “Personalized Search,” which is better than “Personalization”).

Sometimes you will publish two papers on the same topic, often in the form of a preliminary poster or work-in-progress that grows into a conference paper, or a conference paper that grows to a journal article. Avoid using the same title for both publications, since that will make it hard for people to identify the higher quality version. Instead, try to highlight the differences in the name (e.g., poster: How Medical Expertise Influences Web Search Interaction, paper: Characterizing the Influence of Domain Expertise on Web Search Behavior), and reserve the cute, catchy short phrase for the piece that you would like everyone to ultimately refer to.

Although abstracts are longer than titles, like titles they are also surprisingly formulaic, as illustrates by the abstract from The Search Dashbaord: How Reflection and Comparison Impact Search Behavior. An abstract typically starts with a sentence or two that describes the general problem space and why it is important:
Most searchers do not know how to use Web search engines as effectively as possible. This is due, in part, to search engines not providing feedback about how search behavior can be improved.
It then describes the approach taken in the paper to address the problem:
Because feedback is an essential part of learning, we created the Search Dashboard, which provides an interface for reflection on personal search behavior. The Dashboard aggregates and presents an individual’s search history and provides comparisons with that of archetypal expert profiles.
Next comes a brief description of the methodology, including some notion of scale:
Via a five-week study of 90 Search Dash-board users,
Followed by a few important sentences describing the key findings:
we find that users are able to change aspects of their behavior to be more in line with that of the presented expert searchers. We also find that reflection can be beneficial, even without comparison, by changing participants’ views about their own search skills, what is possible with search, and what aspects of their behavior may influence search success.
It is nice to conclude with a sentence about why the findings are important or how they can be used:
Our findings demonstrate a new way for search engines to help users modify their search behavior for positive outcomes.
For both the title and abstract, you should not use jargon, citations, and acronyms, and you should avoid defining new terms. It is also a good idea to avoid special characters or formatting, since both the title and abstract will sometimes be printed in plain text. I screwed up, for example, when I titled a paper “Where’d It Go?”: How People Ask After Lost Web Information, since the punctuation creates many ways for people to get it wrong. On the other hand, be sure to include any keywords you think a person might use when searching for your work, so that the paper will rank highly for relevant queries.

Note that while you want your title and abstract to draw people in, you should be wary of false advertising. Do not overstate the findings presented in the paper or make unsupported claims. The first people to read your paper will be your reviewers, and they are much more likely to reject it if it disappoints their expectations. If you think there is weakness that you will need to convince your reviewers to buy into (such a small study size or a controversial method), mention it as early as possible versus sweeping it under the rug.


This post is the first in a series of posts about the formula for academic papers. The components being discussed are:


  1. This is a great idea for a series, Jaime. I'm in complete agreement about the power of the colon for paper titles, and inventing a short catchphrase for the work. Most of my papers are systems, so the catchphrase usually ends up being the system name.

    One question that sometimes comes up for me: when in the writing process should you write the abstract? Before you write the rest of the paper, or after? I usually wait to write the abstract until the rest of the paper has been mostly or completely drafted, because only then do I know exactly what I've said. :) That's very different than the introduction, which I always write (and iterate on) *before* the rest of the paper.

    I've noticed that students who write the abstract *before* writing the paper often tend to produce a bloated abstract with pretensions for being an introduction, rather than standing alone as a one-paragraph summary of the whole work. And then the introduction suffers, because the student thinks the abstract was the first paragraph of the introduction (which it isn't).

    1. Good points, Rob -- definitely important to be thoughtful about the relationship between the Abstract and Introduction. A related thing that bugs me is when there is text copy-and-pasted between the two sections.

  2. Great post!!!
    (And some brilliantly conceived 'short phrases' as examples)

  3. The formula for writing abstracts, articles is similar in the way you structure because there should be the arguments, examples, conclusion etc. See precis writing to find more about precis writing.