Hierarchically Structured Reinforcement Learning for Topically Coherent Visual Story Generation

TL;DR

This paper introduces a hierarchical reinforcement learning framework for visual storytelling that improves coherence by planning topics at a high level and generating sentences conditioned on these topics, trained end-to-end.

Contribution

The paper presents a novel hierarchical reinforcement learning approach with a two-level decoder for more coherent visual story generation, outperforming flat models.

Findings

Significantly better performance than baseline models.

Improved story coherence and relevance.

Effective end-to-end training of hierarchical structure.

Abstract

We propose a hierarchically structured reinforcement learning approach to address the challenges of planning for generating coherent multi-sentence stories for the visual storytelling task. Within our framework, the task of generating a story given a sequence of images is divided across a two-level hierarchical decoder. The high-level decoder constructs a plan by generating a semantic concept (i.e., topic) for each image in sequence. The low-level decoder generates a sentence for each image using a semantic compositional network, which effectively grounds the sentence generation conditioned on the topic. The two decoders are jointly trained end-to-end using reinforcement learning. We evaluate our model on the visual storytelling (VIST) dataset. Empirical results from both automatic and human evaluations demonstrate that the proposed hierarchically structured reinforced training achieves significantly better performance compared to a strong flat deep reinforcement learning baseline.