Model Criticism for Long-Form Text Generation

Method and contribution

A framework for evaluating high-level text structure by comparing real and generated data in latent spaces defined by explicit generative critics. Experiments study discourse coherence, coreference, and topicality.

The original method behind this guide: diagnostic evaluation of structural distributions, separate from word-level fluency.

Abstract

Language models have demonstrated the ability to generate highly fluent text; however, it remains unclear whether their output retains coherent high-level structure (e.g., story progression). Here, we propose to apply a statistical tool, model criticism in latent space, to evaluate the high-level structure of the generated text. Model criticism compares the distributions between real and generated data in a latent space obtained according to an assumptive generative process. Different generative processes identify specific failure modes of the underlying model. We perform experiments on three representative aspects of high-level discourse—coherence, coreference, and topicality—and find that transformer-based language models are able to capture topical structures but have a harder time maintaining structural coherence or modeling coreference.

Read in context

Evaluating the structure of long-form text →

Citation

Yuntian Deng, Volodymyr Kuleshov, Alexander M. Rush. Model Criticism for Long-Form Text Generation. EMNLP 2022.

@inproceedings{deng-etal-2022-model,
  title = {{Model Criticism for Long-Form Text Generation}},
  author = {Deng, Yuntian and Kuleshov, Volodymyr and Rush, Alexander M.},
  year = {2022},
  booktitle = {EMNLP 2022},
  url = {https://aclanthology.org/2022.emnlp-main.815/},
  eprint = {2210.08444},
  archivePrefix = {arXiv},
  doi = {10.18653/v1/2022.emnlp-main.815}
}