SSAP: Storylines and Sentiment Aware Pre-Trained Model for Story Ending Generation

Yongkang Liu,Qingbao Huang,Jing Li,Linzhang Mo,Yi Cai,Qing Li

DOI: https://doi.org/10.1109/taslp.2022.3145320

2022-01-01

Abstract:As an interesting but under-explored task, story ending generation aims at generating an appropriate ending for an incomplete story. The challenges of the task are to deeply understand the story context, mine the storylines hidden in the story, and generate rational endings in logic and sentiment. Although existing pre-trained approaches have been proven effective to this task, how to learn to generate endings with appropriate plots and sufficient sentimental information still remains a major challenge. One possible reason is that an over reliance on external commonsense knowledge beyond the storylines and sentimental trends information hidden in the story context could lead to generation deviating from the main theme. To address this issue, we propose a two-stage Stroylines and Sentiment Aware Pre-trained model (SSAP) for generating sentimentally relevant story endings. We apply a classifier for discriminating the sentiment of the story, and then employ a pre-trained language model, combining with storylines information, to conditionally generate sentences that match both the logic and sentiment of the story. Automatic and manual evaluations show that, without integrating external knowledge, our model can produce more consistent and diverse story endings than state-of-the-art baselines.