Hindawi / Blog / Blog Post

Latest from our journals

Writing a title and abstract? Get help from AI and big data

Authors | Opinion
Writing a title and abstract? Get help from AI and big data

The title and abstract are key parts of a paper - but they’re not the easiest to write. In this post, Hilde van Zeeland, Chief Applied Linguist at Writefull, shares tips for both, and also shows you how Writefull uses big data and AI to help you write those crucial first words.

Writing a title and abstract? Get help from AI and big data 

A title and abstract should engage, inform, and contribute to a paper’s discoverability. This is something most authors are aware of. Yet writing these first sections of a paper can be a real struggle. 

To learn what makes a good title and abstract, it helps to take published papers as examples. At Writefull, we use this approach to offer language feedback: we apply big data and AI to scientific papers, in order to extract the ‘correct’ (usually the most common) way of writing things. 

In two recent experiments, we applied this approach to take on the title-abstract challenge. Using a dataset of 250,000 Hindawi papers, we created a widget that automatically creates titles, and we analyzed the structure and language of abstracts.

An automatic title generator

When it comes to writing a title, deciding what it should cover often isn’t the biggest challenge; finding the right structure and words is. 

If you want your title to introduce one element only, such as the paper’s subject or method, you can keep its structure simple. For example: "The Emerging Roles of CCN3 Protein in Immune-Related Diseases". However, you may want to combine elements. In this case, try to:

  • Be concise. Select only those elements that matter and use as few words as possible. Articles can often be skipped, and adverbs and adjectives are generally not used.

  • Be informative. Select those elements that are key to your study and that enhance the discoverability of your paper.

To alleviate the struggle of writing a good title, we developed a tool that automatically writes a title based on a paper’s abstract. The widget uses a model that we trained on a set of Hindawi title-abstract pairs. Having processed these examples, the model predicts a suitable title for any new abstract it’s given.

The titles it creates cover the most important information from the abstract, and that follow an established structure. For example, when pasting in the abstract from this paper about the metal contents of lake fish, we get the following titles:

  • Analysis of Fe, Cu, Cd, Cr, and Pb in Fish from Three Lakes Close to Disposal of Industrial Waste in Indonesia

  • Heavy Metals Content of Fish from Three Lakes Close to the Disposal of Industrial Waste

  • The Effects of Heavy Metals Contamination on the Health of Fishes from Three Lakes Near to Industrial Waste

You can have a go at it yourself here: https://x.writefull.com/title-generator/index.html

The structure and language of abstracts

Abstracts are difficult to write in the sense that they should be complete and informative, yet concise. 

The first question is what to include. To decide on this, it may help to imagine yourself explaining your study to a peer. They would be interested in learning:

  • What you studied (topic)

  • Why you studied this (question or problem addressed)

  • How you did it (methods used)

  • What you found (only the key findings!)

  • Why these findings matter (applications or future research)

With this information, your peer would know if the paper is relevant to them. This is also why an abstract usually covers the above points: together, they provide a mini summary of a paper.

Once you’ve decided on the content of your abstract, you can start structuring and writing it. To give you a helping hand with this part, we recently set out to explore what structure abstracts usually follow, and with what words. 

We took Hindawi’s abstracts and extracted the most frequent four-word phrases from them. Next, we divided the abstracts into four parts and counted how often these frequent phrases occurred in each.

This revealed that abstracts follow a clear pattern. The first part usually introduces the topic and aim of the study, with aim of this study being the most frequent phrase. The middle part usually presents the methods, analysis, and results. The last part discusses the significance of the results and study contribution, with frequent phrases such as effectiveness of the proposed X: 

 abstract patterns

Generate the best title

We’re keen to see what the title generator can do on your abstracts. Try it out and share your best generated title (with a screenshot) on Twitter, including @Writefullapp. We’re giving away a free lifetime Writefull subscription to the creator of our favorite title.

This blog post is distributed under the Creative Commons Attribution License (CC-BY). The illustration is by Hindawi and is also CC-BY.

Article of the Year Award: Impactful research contributions of 2022, as selected by our Chief Editors. Discover the winning articles.