Exploring the Relationship Between Diversity and Quality in Ad Text Generation
Yoichi Aoki, Soichiro Murakami, Ukyo Honda, Akihiko Kato
TL;DR
This work systematically investigates how diversity-promoting decoding methods influence ad text quality in Japanese ad generation. Using a five-output setup on the CAMERA dataset and three-shot prompts, the study evaluates diversity via Pairwise-BLEU complements and ad quality via Ad Performance, Ad Consistency, and Ad Acceptability. The findings reveal a trade-off between diversity and ad quality across decoding strategies, with beam-search variants and multi-model ensembles offering different strength profiles; all-at-once strategies tend to increase diversity at the expense of quality. The results inform practical deployment of diversity methods in ad text generation and point to ensemble approaches as a viable path to sustain quality while expanding diversity.
Abstract
In natural language generation for advertising, creating diverse and engaging ad texts is crucial for capturing a broad audience and avoiding advertising fatigue. Regardless of the importance of diversity, the impact of the diversity-enhancing methods in ad text generation -- mainly tested on tasks such as summarization and machine translation -- has not been thoroughly explored. Ad text generation significantly differs from these tasks owing to the text style and requirements. This research explores the relationship between diversity and ad quality in ad text generation by considering multiple factors, such as diversity-enhancing methods, their hyperparameters, input-output formats, and the models.