Friday 21 March 2025
The quest for more effective advertising has led researchers to explore the intricacies of human preference when it comes to online ads. Recently, a team of scientists delved into the world of natural language processing to investigate how linguistic features influence our attraction to ad texts.
By collecting over 1,200 pairs of paraphrased ad texts, the researchers aimed to identify the key factors that make an ad text more attractive to potential customers. The study focused on Japanese ads, which posed unique challenges due to the complexities of the language.
The team employed a combination of human evaluation and machine learning algorithms to assess the attractiveness of each ad pair. Ten judges, including professionals with experience in advertising creation, evaluated the pairs based on criteria such as memorability, eye-catchingness, and ease of understanding. Meanwhile, artificial intelligence models like GPT-3.5 and Llama2 generated paraphrased candidates from a given ad text.
The results revealed that certain linguistic features played a significant role in determining an ad’s attractiveness. For instance, ad texts with more content words – such as nouns, verbs, and adjectives – were found to be more informative and attractive. Additionally, the use of bracket symbols like 【】 or 「」was also associated with improved attractiveness.
Furthermore, the study showed that the specificity of an ad text can have a substantial impact on its appeal. Ad texts that conveyed unique information or emphasized specific benefits were deemed more attractive than those with generic or vague content.
Interestingly, the researchers discovered that the way an ad text is structured can also influence our perception of it. Dependency parsing, which analyzes the grammatical relationships between words in a sentence, revealed that ad texts with longer and more complex dependency trees were often rated as more attractive.
The findings of this study have significant implications for advertisers seeking to create effective online ads. By incorporating linguistic features like content words, bracket symbols, specificity, and structural complexity into their ad campaigns, businesses can increase the likelihood of attracting potential customers.
Moreover, the development of machine learning algorithms capable of generating paraphrased ad texts has the potential to revolutionize the advertising industry. These models could be used to create a wide range of ad variations, allowing advertisers to quickly test different approaches and optimize their campaigns for maximum effectiveness.
In short, this study demonstrates the importance of linguistic features in shaping our perception of online ads.
Cite this article: “The Role of Linguistic Features in Attracting Customers to Online Ads”, The Science Archive, 2025.
Linguistic Features, Advertising, Attractiveness, Language Processing, Machine Learning, Paraphrasing, Japanese Ads, Content Words, Bracket Symbols, Specificity
