[2H6-GS-9] Natural language processing, information retrieval: Document generation

[2H6-GS-9-05] Generating Advertising Copies from Keywords by Using Multiple Language Models

〇HAODA ZHANG¹, Hidetaka Kamigaito¹, Hiroya Takamura^1,2, Manabu Okumura¹ (1. Tokyo Institute of Technology, 2. National Institute of Advanced Industrial Science and Technology)

Keywords:Sentence Generation, Language Model

In making advertising copies, it is important how much they can attract people. However, those copy sentences are made manually nowadays, and it costs a lot. In this paper, we propose a new approach that can generate copy sentences from keywords by extending the existing approach and using multiple language models. It allows us to generate sentences where keywords are put in the beginning and whose lengths are in the specified range. We show that it can generate better sentences by using multiple language models than only by using one.