大規模言語モデルおよび定量的ペルソナ作成手法による顧客ストーリーの生成

抄録

Targeting customers in the asset formation domain necessitates analysis that considers both current customer attributes and future attribute changes. The authors have proposed a simulation-based quantitative persona creation method, but it is limited to establishing persona skeletons, and creating customer stories with additional narratives remains a challenge. Our previous method involves a cross comparison of several skeletons based on a combination of current customer attributes, the possibility of future attribute changes, and the assumed effects of measures, and thus there are limitations to conventional manual story creation. To address this, our study proposes a methodology to generate customer stories based on skeletons mechanically and semi-automatically, using a large language model. The main findings of this research are as follows: our methodology could 1) generate a set of stories with a certain level of consistency in terms of structure, 2) write stories according to the attributes of each skeleton, 3) generate logically coherent stories, incorporating future attribute changes, and 4) provide suggestions to foster awareness among stakeholders involved in the final phase of persona creation. The proposed method is applicable not only to persona creation but also to lifecycle management, such as version upgrades, and is expected to find industrial applications.