Mars: Semantic-aware Contrastive Learning for End-to-End Task-Oriented Dialog

Sun, Hao; Bao, Junwei Lucas; Wu, Youzheng; He, Xiaodong

doi:10.48550/arxiv.2210.08917

Cited by 4 publications

(6 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We observe that in most cases, error originates from incorrectly generated system action and dialog state. This hints at a direction for further improvement in lines of making dialog state and system action generation more robust (Sun et al, 2022b). 12354…”

Section: O Krls Error Examplesmentioning

confidence: 98%

“…We present this result in Table 7, and found that +DST improved the overall score by nearly 4 points, and +Both further improved the overall score by 14.5 points, almost reaching the performance of the training dataset 4 . This shows that much error remains in the DST and system act generation process, so the overall performance can further increase if techniques to separately improve DST and system act generation (e.g., Sun et al (2022b)) can be combined with KRLS. We leave this for future work.…”

Section: Error Analysismentioning

confidence: 99%

See 1 more Smart Citation

KRLS: Improving End-to-End Response Generation in Task Oriented Dialog with Reinforced Keywords Learning

Yu,

Wu,

Qian

et al. 2023

Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing

View full text Add to dashboard Cite

In task-oriented dialogs (TOD), reinforcement learning (RL) algorithms train a model to directly optimize response for task-related metrics. However, RL needs to perform exploration, which can be time-consuming due to the slow auto-regressive sequence generation process. We investigate an approach to create a more efficient RL-based algorithm to improve TOD performance in an offline setting. First, we use a faster generation procedure that samples from independent next-word distributions after training the language model (LM) with supervised learning. We then introduce a finegrained reward function to help the model focus on learning key information in a dialog, by measuring the importance and semantic closeness of each generated token. Experiments on the MultiWoZ dataset show our new training algorithm, Keywords Reinforcement Learning with Next-word Sampling (KRLS), achieves state-of-the-art performance on the end-to-end response generation task, with a 15% training time reduction compared to a standard RL algorithm using auto-regressive generation 1 .

show abstract

Section: O Krls Error Examplesmentioning

confidence: 98%

Section: Error Analysismentioning

confidence: 99%

KRLS: Improving End-to-End Response Generation in Task Oriented Dialog with Reinforced Keywords Learning

Yu,

Wu,

Qian

et al. 2023

Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing

View full text Add to dashboard Cite

show abstract

“…Building task-oriented dialog (TOD) systems has been a long-standing challenge in artificial intelligence. The prevailing approach for creating task bots (Hosseini-Asl et al, 2020;Peng et al, 2021a;Sun et al, 2022) is to fine-tune pre-trained language models (PLMs), such as T5 (Raffel et al, 2020) and GPT-2 (Radford et al, 2019). Despite their great success, developing and maintaining such task bots generally requires adequate annotated data and extensive fine-tuning/re-training.…”

Section: Introductionmentioning

confidence: 99%

SGP-TOD: Building Task Bots Effortlessly via Schema-Guided LLM Prompting

Zhang,

Peng,

et al. 2023

Findings of the Association for Computational Linguistics: EMNLP 2023

View full text Add to dashboard Cite

Building and maintaining end-to-end task bots using minimal human effort is a longstanding challenge in dialog research. In this work, we introduce SGP-TOD, S ¯chema-G ¯uided P ¯rompting for building T ¯ask-O ¯riented D ¯ialog systems effortlessly based on large language models (LLMs). Utilizing the predefined task schema, i.e., belief instruction and dialog policy, we instruct fixed LLMs to generate appropriate responses on novel tasks, without the need for training data. Specifically, SGP-TOD comprises three components: an LLM for interacting with users, a Dialog State Tracking (DST) Prompter to aid the LLM in tracking dialog states with the given belief instruction, and a Policy Prompter to direct the LLM to generate proper responses adhering to the provided dialog policy. Experimental results on Multiwoz, RADDLE, and STAR datasets show that our training-free strategy, SGP-TOD, yields stateof-the-art (SOTA) zero-shot performance, significantly surpassing the few-shot approaches. In a domain-extension setting, SGP-TOD aptly adapts to new functionalities by merely adding supplementary schema rules. We make our code and data publicly available. 1

show abstract

“…We evaluate both end-to-end and policy optimization settings. This includes UBAR (Nekvinda and Dusek, 2021), PPTOD (Su et 2022), RSTOD (Cholakov and Kolev, 2022), BORT (Sun et al, 2022a), MTTOD (Lee, 2021), HDNO (Wang et al, 2020a), GALAXY , MarCO (Wang et al, 2020b), Mars (Sun et al, 2022b), and KRLS . To obtain database search results in the end-to-end setting, we use MTTOD's dialogue state tracker, which is trained jointly during fine-tuning.…”

Section: Experiments Setupmentioning

confidence: 99%

“…Response generation is an important task in taskoriented dialogue systems. There have been many previous approaches (Hosseini-Asl et al, 2020;Gu et al, 2021;Su et al, 2022;Sun et al, 2022b;Wu et al, 2023) proposed to improve the task-oriented dialogue systems. One direction is the use of dialogue act annotations to improve the quality of responses in task-oriented dialogue systems.…”

Section: Introductionmentioning

confidence: 99%

Learning to memorize in neural task-oriented dialogue systems

Wu¹

View full text Add to dashboard Cite

Dialogue act annotations are important to improve response generation quality in taskoriented dialogue systems. However, it can be challenging to use dialogue acts to control response generation in a generalizable way because different datasets and tasks may have incompatible annotations. While alternative methods that utilize latent action spaces or reinforcement learning do not require explicit annotations, they may lack interpretability or face difficulties defining task-specific rewards. In this work, we present a novel end-to-end latent dialogue act model (DiactTOD) that represents dialogue acts in a latent space. Diact-TOD, when pre-trained on a large corpus, is able to predict and control dialogue acts to generate controllable responses using these latent representations in a zero-shot fashion. Our approach demonstrates state-of-the-art performance across a wide range of experimental settings on the MultiWOZ dataset, including zeroshot, few-shot, and full data fine-tuning with both end-to-end and policy optimization configurations.

show abstract

Mars: Semantic-aware Contrastive Learning for End-to-End Task-Oriented Dialog

Cited by 4 publications

References 22 publications

KRLS: Improving End-to-End Response Generation in Task Oriented Dialog with Reinforced Keywords Learning

KRLS: Improving End-to-End Response Generation in Task Oriented Dialog with Reinforced Keywords Learning

SGP-TOD: Building Task Bots Effortlessly via Schema-Guided LLM Prompting

Learning to memorize in neural task-oriented dialogue systems

Contact Info

Product

Resources

About