Do Massively Pretrained Language Models Make Better Storytellers?

TL;DR

This paper evaluates whether large pretrained language models like GPT2-117 produce better storytelling than specialized models, finding they excel in context sensitivity but still face issues like repetition and lack of diversity.

Contribution

It provides a detailed comparison of pretrained and specialized story generation models using automatic metrics, highlighting strengths and limitations of large pretrained models.

Findings

GPT2-117 conditions more strongly on context

More sensitive to event order and uses more unusual words

Still produces repetitive and less diverse text

Abstract

Large neural language models trained on massive amounts of text have emerged as a formidable strategy for Natural Language Understanding tasks. However, the strength of these models as Natural Language Generators is less clear. Though anecdotal evidence suggests that these models generate better quality text, there has been no detailed study characterizing their generation abilities. In this work, we compare the performance of an extensively pretrained model, OpenAI GPT2-117 (Radford et al., 2019), to a state-of-the-art neural story generation model (Fan et al., 2018). By evaluating the generated text across a wide variety of automatic metrics, we characterize the ways in which pretrained models do, and do not, make better storytellers. We find that although GPT2-117 conditions more strongly on context, is more sensitive to ordering of events, and uses more unusual words, it is just as likely to produce repetitive and under-diverse text when using likelihood-maximizing decoding algorithms.