Hafez: an Interactive Poetry Generation System

By Marjan Ghazvininejad, Xing Shi, Jay Priyadarshi, Kevin Knight. 2017

169 citations recorded by the source corpus.

Hafez is an automatic poetry generation system that integrates a Recurrent Neural Network (RNN) with a Finite State Accep-tor (FSA). It generates sonnets given arbitrary topics. Furthermore, Hafez enables users to revise and polish generated poems by adjusting various style conﬁgurations. Experiments demonstrate that such “pol-ish” mechanisms consider the user’s intention and lead to a better poem. For evaluation, we build a web interface where users can rate the quality of each poem from 1 to 5 stars. We also speed up the whole system by a factor of 10, via vocabulary pruning and GPU computation, so that adequate feedback can be collected at a fast pace. Based on such feedback, the system learns to adjust its parameters to improve poetry quality.

Subfields: Artificial Intelligence, Control and Systems Engineering, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Control and Systems Engineering

In Natural Language Processing

Citations in the corpus

Open-Domain Conversational Agents: Current Progress, Open Problems, and Future Directions (2020, methodology): “, 2018) and weighted decoding (Ghazvininejad et al., 2017; Baheti et al., 2018).”
GeDi: Generative Discriminator Guided Sequence Generation (2020, background): “…et al., 2021; Madotto et al., 2020; Keskar et al., 2019; Ziegler et al., 2019; Rajani et al., 2019; Fan et al., 2018; Ficler and Goldberg, 2017; Yu et al., 2017; Hu et al., 2017) or using a discriminator to guide decoding (Ghazvininejad et al., 2017; Holtzman et al., 2018; Dathathri et al., 2020).”
Plug and Play Language Models: A Simple Approach to Controlled Text Generation (2019, methodology): “, 2019); and WD: a weighted decoding baseline in which the B model’s outputs are weighted directly toward maximizing p(a|x) (Ghazvininejad et al., 2017); see Section S6 for details.”
Decision Transformer: Reinforcement Learning via Sequence Modeling (2021, background): “Class-conditional language models can also be used to learn disciminators to guide generation (Dathathri et al., 2019; Ghazvininejad et al., 2017; Holtzman et al., 2018; Krause et al., 2020).”
What makes a good conversation? How controllable attributes affect human judgments (2019, methodology): “Given these constraints, we study two con-trol methods: conditional training (Fan et al., 2018; Kikuchi et al., 2016; Peng et al., 2018) and weighted decoding (Ghazvininejad et al., 2017).”
Towards Controllable Biases in Language Generation (2020, methodology): “ applied control to various components in a model pipeline. Keskar et al. (2019) present a large language model that learns during the training procedure to control for style and other speciﬁc tasks. Ghazvininejad et al. (2017) use weighted decoding to control the style of generated poems. Dathathri”
Towards Controllable Story Generation (2018, background): “…on controllable generation of images (Chen et al., 2016; Siddharth et al., 2017; Lample et al., 2017), dialog responses (Wang et al., 2017), poems (Ghazvininejad et al., 2017), and different styles of text (Hu et al., 2017; Ficler and Goldberg, 2017; Shen et al., 2017; Fu et al., 2017). people…”
Don’t Say That! Making Inconsistent Dialogue Unlikely with Unlikelihood Training (2019, background): “In terms of improving repetition and speciﬁcity, a recent alternative approach is that of control (Fan et al., 2018; Ficler and Goldberg, 2017; Ghazvininejad et al., 2017; See et al., 2019).”
FUDGE: Controlled Text Generation With Future Discriminators (2021, methodology): “, G’s output logits), and operates directly on these logits (Ghazvininejad et al., 2017; Holtzman et al., 2018; Cohn-Gordon et al., 2018; Shen et al., 2019).”
PowerTransformer: Unsupervised Controllable Revision for Biased Language Correction (2020, methodology): “Using machine-in-the-loop writing systems (e.g., Ghazvininejad et al., 2016, 2017; Clark et al., 2018, Textio 2 ), models trained on this task could help authors write news, stories, or movies that portray characters in less biased ways, and thereby help mitigate the negative effects of…”
Detect and Perturb: Neutral Rewriting of Biased and Sensitive Text via Gradient-based Decoding (2021, unknown): “• Weighed Decoding (WD): a decoding method (Ghazvininejad et al., 2017) by reducing the generation probability of detected sensitive tokens to a hyperparameter α (we set α = 0 .”
Story Centaur: Large Language Model Few Shot Learning as a Creative Writing Tool (2021, background): “It is certainly the case that recent successful work in rhyming metered poetry generation has needed to resort to ﬁxed rhyme words and syllable counting(Ghazvininejad et al., 2017).”
SynthBio: A Case Study in Human-AI Collaborative Curation of Text Datasets (2021, background): “This follows a rich line of research in human-model collaborative interfaces, in which models generate draft outputs that humans edit [22, 10, 28, 13, 24, 23].”
Choose Your Own Adventure: Paired Suggestions in Collaborative Writing for Evaluating Story Generation Models (2021, background): “Collaborative writing systems have been developed in domains like poetry (Ghazvininejad et al., 2017), slogans (Clark et al., 2018), and stories (Roemmele and Gordon, 2015; Goldfarb-Tarrant et al., 2019; Akoury et al., 2020).”
Offline RL for Natural Language Generation with Implicit Language Q Learning (2022, methodology): “We compare ILQL to: (1) ILQL (utterance): a per-utterance adaptation of ILQL that is nearly identical to ILQL, except that it removes the conservatism loss term and performs Bellman backups at the utterance level instead of the token level; (2) SARSA (utterance): a per-utterance version of SARSA, wh”
Towards a Science of Human-AI Decision Making: A Survey of Empirical Studies (2021, methodology): “Deep learning models Convolution Neural Networks [3, 27], Recurrent Neural Networks [26, 60], BERT [80], RoBERTa [13], VQA model (hybrid LSTM and CNN) [115], not specified [23, 24, 30, 48, 50, 52, 62, 73, 86, 87, 107, 110, 152]”
Policy-Driven Neural Response Generation for Knowledge-Grounded Dialog Systems (2020, methodology): “Weighted decoding (See et al., 2019) instead uses token-level features that are controllable (Ghazvininejad et al., 2017; Baheti et al., 2018) and supplements the scores from the de-coder model output with these features.”
CoCon: A Self-Supervised Approach for Controlled Text Generation (2020, background): “Weighted decoding [9, 12] seeks to control the output text token by upweighting the probabilities of targeted words during the decoding step but has been shown to produce incoherent text [25].”
Plug-and-Play Conversational Models (2020, methodology): “…plug-and-play settings such as: DG : DialoGPT proposed by Zhang et al. (2019); WD : DialoGPT plus a word level weight-decoding schema as in (Ghazvininejad et al., 2017; See et al., 2019); PP : DialoGPT plus PPLM (Dathathri et al., 2019), as explained in Section 3.1; AD : DialoGPT with one…”
Nine Potential Pitfalls when Designing Human-AI Co-Creative Systems (2021, background): “org) also entered specifically artistic domains, including visual art [7], creative writing and poetry [8, 9].”