论文信息 - Improving Language Generation with Sentence Coherence Objective

Improving Language Generation with Sentence Coherence Objective

Conditional story generation and contextual text continuation have become increasingly popular topics in NLP community. Existing models are often prone to output paragraphs of texts that gradually diverge from the given prompt. Although the generated text may have a reasonable perplexity and diversity, it could easily be identified by human as gibberish. The goal of our project is to improve the coherence and consistency across sentences in a language-generation model. We aim to solve this issue by first training a sentence pair coherence classifier with GPT-2 pretrained model, and then co-train the GPT-2 language model with this new coherence objective using a method analogous to the REINFORCE algorithm. This fine-tuned language model is able to generate lengthy paragraph conditioned on a given topic without diverging too much. The simplicity of this model allows it to be applicable to a variety of underlying language model architecture since it only modifies the final layer of the pre-trained model.

[1] Alexander M. Rush,et al. Encoder-Agnostic Adaptation for Conditional Language Generation , 2019, ArXiv.

[2] Douglas Eck,et al. Generating Music by Fine-Tuning Recurrent Neural Networks with Reinforcement Learning , 2016 .

[3] Alec Radford,et al. Improving Language Understanding by Generative Pre-Training , 2018 .

[4] Ming-Wei Chang,et al. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding , 2019, NAACL.

[5] Ronald J. Williams,et al. Simple Statistical Gradient-Following Algorithms for Connectionist Reinforcement Learning , 2004, Machine Learning.

[6] Ilya Sutskever,et al. Language Models are Unsupervised Multitask Learners , 2019 .

[7] Yann Dauphin,et al. Hierarchical Neural Story Generation , 2018, ACL.

[8] Mina Lee,et al. Learning Autocomplete Systems as a Communication Game , 2019, ArXiv.

[9] Oriol Vinyals,et al. Representation Learning with Contrastive Predictive Coding , 2018, ArXiv.

[10] Sanja Fidler,et al. Skip-Thought Vectors , 2015, NIPS.

[11] Melissa Roemmele,et al. Writing Stories with Help from Recurrent Neural Networks , 2016, AAAI.