Directed Beam Search: Plug-and-Play Lexically Constrained Language Generation

Pascual, Damián; Egressy, Béni; Florian, Bolli,; Wattenhofer, Roger

doi:10.48550/arxiv.2012.15416

Cited by 6 publications

(14 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Previous works proposed beam search variants specifically for lexically constrained decoding (Hokamp & Liu, 2017;Pascual et al, 2020;Lu et al, 2021) which enforce constraints during search in a discrete space. Recent works consider constraint satisfaction by adjusting vocabulary distributions using an additional discriminator or LM (Dathathri et al, 2019;Krause et al, 2021).…”

Section: Related Workmentioning

confidence: 99%

COLD Decoding: Energy-based Constrained Text Generation with Langevin Dynamics

Qin¹,

Welleck²,

Khashabi³

et al. 2022

Preprint

View full text Add to dashboard Cite

Many applications of text generation require incorporating different constraints to control the semantics or style of generated text. These constraints can be hard (e.g., ensuring certain keywords are included in the output) and soft (e.g., contextualizing the output with the left-or right-hand context). In this paper, we present Energy-based Constrained Decoding with Langevin Dynamics (COLD), a decoding framework which unifies constrained generation as specifying constraints through an energy function, then performing efficient differentiable reasoning over the constraints through gradient-based sampling. COLD decoding is a flexible framework that can be applied directly to off-the-shelf left-to-right language models without the need for any task-specific finetuning, as demonstrated through three challenging text generation applications: lexically-constrained generation, abductive reasoning, and counterfactual reasoning. Our experiments on these constrained generation tasks point to the effectiveness of our approach, both in terms of automatic and human evaluation. 1

show abstract

Section: Related Workmentioning

confidence: 99%

COLD Decoding: Energy-based Constrained Text Generation with Langevin Dynamics

Qin¹,

Welleck²,

Khashabi³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Black-box approaches including "Prompt Engineering" that aim to change the prompts fed into the base LM at inference time (Wallace et al, 2019;Li and Liang, 2021). Guided generation targets at building a controllable "guiding" model that shifts the output from base LM at inference time (Krause et al, 2020;Pascual et al, 2020).…”

Section: Plug-and-play Conditional Generationmentioning

confidence: 99%

“…We seek "plug-and-play" approaches to controllable text generation wherein new language models can be slotted into existing generative systems; new language models are being developed and it becomes intractable to update and retrain controlled generation architectures. Plug-and-play techniques such as (Krause et al, 2020;Pascual et al, 2020) aim to only intervene with the outputs-a vector of logits-of a generative language model. This becomes especially important as the latest iteration of very large pre-trained language models such as GPT-3 (Brown et al, 2020) restrict access to the hidden states and layer weights of models.…”

Section: Introductionmentioning

confidence: 99%

Plug-and-Blend: A Framework for Controllable Story Generation with Blended Control Codes

Lin¹,

Riedl²

2021

Preprint

View full text Add to dashboard Cite

We describe a Plug-and-Play controllable language generation framework, Plug-and-Blend, that allows a human user to input multiple control codes (topics). In the context of automated story generation, this allows a human user lose or fine grained control of the topics that will appear in the generated story, and can even allow for overlapping, blended topics. We show that our framework, working with different generation models, controls the generation towards given continuous-weighted control codes while keeping the generated sentences fluent, demonstrating strong blending capability.

show abstract

“…The goal is to generate a given set of control words in the responses of one of the speakers (agent or customer). Naive constrained generation approaches (Pascual et al, 2020;Miao et al, 2019) use methods like beam search and stochastic search to force the generation of these control words for short-term control, where control words need to appear in a single utterance or phrase. Because they do not consider the future, these approaches may generate the words all at once in a single response or not generate them at natural places in the conversation (Figure 1, left).…”

Section: Introductionmentioning

confidence: 99%

“…To alleviate this issue, we retrieve similar conversations from training and condition on them during generation. We first identify similar neighbors using a kNN-based approach and then guide the language model towards generating similar responses, inspired by plug-andplay methods (Madotto et al, 2021;Dathathri et al, 2019;Pascual et al, 2020). The motivation for this is that retrieved conversations guide the model to generate the control words at more natural points in the conversation.…”

Section: Introductionmentioning

confidence: 99%

Long-term Control for Dialogue Generation: Methods and Evaluation

Ramakrishnan¹,

Narangodage²,

Schilman³

et al. 2022

Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Langua

View full text Add to dashboard Cite

Current approaches for controlling dialogue response generation are primarily focused on high-level attributes like style, sentiment, or topic. In this work, we focus on constrained long-term dialogue generation, which involves more fine-grained control and requires a given set of control words to appear in generated responses. This setting requires a model to not only consider the generation of these control words in the immediate context, but also produce utterances that will encourage the generation of the words at some time in the (possibly distant) future. We define the problem of constrained long-term control for dialogue generation, identify gaps in current methods for evaluation, and propose new metrics that better measure long-term control. We also propose a retrieval-augmented method that improves performance of long-term controlled generation via logit modification techniques. We show through experiments on three task-oriented dialogue datasets that our metrics better assess dialogue control relative to current alternatives and that our method outperforms state-of-theart constrained generation baselines. 1

show abstract

Directed Beam Search: Plug-and-Play Lexically Constrained Language Generation

Cited by 6 publications

References 0 publications

COLD Decoding: Energy-based Constrained Text Generation with Langevin Dynamics

COLD Decoding: Energy-based Constrained Text Generation with Langevin Dynamics

Plug-and-Blend: A Framework for Controllable Story Generation with Blended Control Codes

Long-term Control for Dialogue Generation: Methods and Evaluation

Contact Info

Product

Resources

About