Unit 5: Tips for Writers

Writing the abstract before the full paper forces you to speculate about the final findings and structure. Writing it last — after the paper is complete — means you can extract exact sentences from the body of the paper and are certain of what you have actually found.

Exception: Some conferences require a submission of the abstract before the full paper is due. In this case, treat the abstract as a draft and revise it once the paper is complete.

The summary sentence at the end of your introduction, the opening sentence of your method section, your primary result sentence, and the conclusion of your discussion section are often the best candidates for inclusion in the abstract. Assemble these into a draft, then refine the connections between them.

This approach ensures that the abstract is consistent with the paper — a common problem when the abstract is written from memory rather than from the text itself.

Even if you are writing a traditional (headingless) abstract, begin by using the headings Introduction, Purpose, Method, Results, and Discussion as scaffolding. Using headings makes it easy to see the balance of the abstract — whether any section is disproportionately long or short.

Once you are satisfied with the content, remove the headings (unless you are writing a structured abstract) and refine the prose to flow naturally.

The sections that deserve the most emphasis should be the longest. In computer science, the Method and Results sections are typically the most important and therefore the longest. An abstract that spends 60% of its words on introduction and only 10% on results sends the wrong signal to reviewers.

If you are using a structured format, consider tailoring the headings to better reflect your research rather than using generic labels. For example, instead of Method, you might write Problem Breakdown, System Architecture, or Evaluation Protocol. More specific headings signal domain expertise.

Do not assume reviewers will identify what is new about your work. State it explicitly. Common signals include:

• "We are the first to…"
• "Unlike previous approaches, our method…"
• "To address the gap in X, we propose Y…"

Even if the novelty feels obvious to you, a reviewer skimming dozens of abstracts needs it stated clearly.

If the significance of your work is not self-evident, state it. This is particularly important for applied research that addresses a real-world problem. Phrases like "with direct implications for…" or "providing a practical tool for…" help reviewers understand the value of the contribution beyond the technical result.

Write the abstract two or three times using different move patterns (IMRD, IPMRD, RM) and compare the results. The version that communicates the contribution most clearly and compellingly — using the fewest words — is usually the best choice.

Abstract verbosity is a common problem. Every word in an abstract should carry meaning. Common sources of unnecessary words include:

• Hedges that add nothing: "it is worth noting that…"
• Meta-commentary: "this paper presents a study which investigates…" → "we investigate…"
• Redundant pairs: "analyse and examine" → "examine"

Once you are satisfied with the content and structure, remove the scaffold headings and ensure the abstract reads as fluent prose. Use transitional phrases to connect the moves: "To this end,…", "Our results show that…", "These findings suggest…"

Apply the five-filter approach from the Error-Free Research Writing course to check for common language errors in your abstract:

1. Accuracy: Are all facts, numbers, and claims correct and consistent with the paper?
2. Brevity: Are there redundant words or phrases that can be removed?
3. Clarity: Is every term specific and unambiguous?
4. Objectivity: Is the writing depersonalised and free of emotive language?
5. Formality: Is the register formal and consistent with the target venue?

Ask a colleague who is not closely involved in your research to read your abstract. Immediately after reading, ask them: "What was the main point of this research?" and "What did the researchers find?" If they cannot answer these questions, the abstract needs revision.

This test is the single most effective way to check whether the abstract communicates its key contribution clearly.

The title should summarise the most important content of the abstract — which is itself a summary of the paper. A good title names the key method, result, or contribution without being so specific that it excludes readers. Vague titles like "A Study of Machine Learning" are less effective than specific titles like "Curriculum Learning Improves Medical Text Summarisation by 23%".

Many conferences and journals specify a maximum word count, a required structure, or specific headings for abstracts. Before finalising, verify that your abstract adheres to the submission guidelines. Non-compliance may result in desk rejection — the paper being returned without review.