# From Plots to Endings: A Reinforced Pointer Generator for Story Ending Generation

## TL;DR

This paper presents a novel reinforcement learning-based framework for story ending generation that improves coherence and relevance by combining pointer-generator networks with a reward-driven fine-tuning approach.

## Contribution

It introduces a new task and a combined generator-reward framework that enhances story ending coherence and relevance over existing models.

## Key findings

- Outperforms baseline by 15.75% in CIDEr score
- Achieves 13.57% higher in story consistency
- Effective handling of OOV and repetitive words

## Abstract

We introduce a new task named Story Ending Generation (SEG), whic-h aims at generating a coherent story ending from a sequence of story plot. Wepropose a framework consisting of a Generator and a Reward Manager for thistask. The Generator follows the pointer-generator network with coverage mech-anism to deal with out-of-vocabulary (OOV) and repetitive words. Moreover, amixed loss method is introduced to enable the Generator to produce story endingsof high semantic relevance with story plots. In the Reward Manager, the rewardis computed to fine-tune the Generator with policy-gradient reinforcement learn-ing (PGRL). We conduct experiments on the recently-introduced ROCStoriesCorpus. We evaluate our model in both automatic evaluation and human evalua-tion. Experimental results show that our model exceeds the sequence-to-sequencebaseline model by 15.75% and 13.57% in terms of CIDEr and consistency scorerespectively.