GlobalWoZ: Globalizing MultiWoZ to Develop Multilingual Task-Oriented Dialogue Systems

Ding, Bosheng; Hu, Junjie; Mahani, Sharifah Aljunied; Joty, Shafiq; Si, Luo; Chen, Miao

doi:10.18653/v1/2022.acl-long.115

Cited by 18 publications

(27 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Two concurrent works also derive multilingual datasets from MultiWOZ (Ding et al, 2021;Zuo et al, 2021), with different strategies and properties, discussed in what follows.…”

Section: Comparison With Concurrent Workmentioning

confidence: 99%

“…GlobalWOZ (Ding et al, 2021) encompasses Chinese, Indonesian, and Spanish datasets. The authors first create templates from dialog utterances by replacing slot-value strings in the utterances with the slot type and value index (e.g., ".…”

Section: Comparison With Concurrent Workmentioning

confidence: 99%

“…Compared to the products of concurrent efforts that derive multilingual datasets from English Mul-tiWOZ (Ding et al, 2021;Zuo et al, 2021), our MULTI 2 WOZ is: (1) much larger -we translate all dialogs from development and test portions of the English MultiWOZ (in total 2,000 dialogs containing the total of 29.5K utterances); (2) much more reliable -complete dialogs, i.e., utterances as well as slot-values, have been manually translated (without resorting to error-prone heuristics), and the quality of translations has been validated through quality control steps; and (3) parallel -the same set of dialogs has been translated to all target languages, enabling the direct comparison of the performance of multilingual models and crosslingual transfer approaches across languages.…”

Section: Introductionmentioning

confidence: 99%

“…In this work, we introduce MULTI 2 WOZ, a new multilingual multidomain TOD dataset, derived from the wellestablished English dataset MULTIWOZ, that spans four typologically diverse languages: Chinese, German, Arabic, and Russian. In contrast to concurrent efforts (Ding et al, 2021;Zuo et al, 2021), MULTI 2 WOZ contains goldstandard dialogs in target languages that are directly comparable with development and test portions of the English dataset, enabling reliable and comparative estimates of cross-lingual transfer performance for TOD. We then introduce a new framework for multilingual conversational specialization of pretrained language models (PrLMs) that aims to facilitate crosslingual transfer for arbitrary downstream TOD tasks.…”

mentioning

confidence: 99%

See 3 more Smart Citations

Multi2WOZ: A Robust Multilingual Dataset and Conversational Pretraining for Task-Oriented Dialog

Hung¹,

Lauscher²,

Vulić³

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

Research on (multi-domain) task-oriented dialog (TOD) has predominantly focused on the English language, primarily due to the shortage of robust TOD datasets in other languages, preventing the systematic investigation of cross-lingual transfer for this crucial NLP application area. In this work, we introduce MULTI 2 WOZ, a new multilingual multidomain TOD dataset, derived from the wellestablished English dataset MULTIWOZ, that spans four typologically diverse languages: Chinese, German, Arabic, and Russian. In contrast to concurrent efforts (Ding et al., 2021;Zuo et al., 2021), MULTI 2 WOZ contains goldstandard dialogs in target languages that are directly comparable with development and test portions of the English dataset, enabling reliable and comparative estimates of cross-lingual transfer performance for TOD. We then introduce a new framework for multilingual conversational specialization of pretrained language models (PrLMs) that aims to facilitate crosslingual transfer for arbitrary downstream TOD tasks. Using such conversational PrLMs specialized for concrete target languages, we systematically benchmark a number of zero-shot and few-shot cross-lingual transfer approaches on two standard TOD tasks: Dialog State Tracking and Response Retrieval. Our experiments show that, in most setups, the best performance entails the combination of (i) conversational specialization in the target language and (ii) few-shot transfer for the concrete TOD task. Most importantly, we show that our conversational specialization in the target language allows for an exceptionally sample-efficient fewshot transfer for downstream TOD tasks.

show abstract

“…Two concurrent works also derive multilingual datasets from MultiWOZ (Ding et al, 2021;Zuo et al, 2021), with different strategies and properties, discussed in what follows.…”

Section: Comparison With Concurrent Workmentioning

confidence: 99%

Section: Comparison With Concurrent Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

mentioning

confidence: 99%

See 2 more Smart Citations

Multi2WOZ: A Robust Multilingual Dataset and Conversational Pretraining for Task-Oriented Dialog

Hung¹,

Lauscher²,

Vulić³

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

show abstract

“…Language planning [17,28,55,63,64] is a task that aims to decompose a specific task into a sequence of temporally extended procedural steps. Such ability is a necessity for a wide range of downstream applications, such as household robots [1,2,32,34,44] and conversational assistants [3,9,11,19,58].…”

Section: Introductionmentioning

confidence: 99%

Neuro-Symbolic Causal Language Planning with Commonsense Prompting

Lu¹,

Feng²,

Zhu³

et al. 2022

Preprint

View full text Add to dashboard Cite

Language planning aims to implement complex high-level goals by decomposition into sequential simpler low-level steps. Such procedural reasoning ability is essential for applications such as household robots and virtual assistants. Although language planning is a basic skill set for humans in daily life, it remains a challenge for large language models (LLMs) that lack deep-level commonsense knowledge in the real world. Previous methods require either manual exemplars or annotated programs to acquire such ability from LLMs. In contrast, this paper proposes Neuro-Symbolic Causal LAnguage Planner (CLAP) that elicits procedural knowledge from the LLMs with commonsense-infused prompting. Pre-trained knowledge in LLMs is essentially an unobserved confounder that causes spurious correlations between tasks and action plans. Through the lens of a Structural Causal Model (SCM), we propose an effective strategy in CLAP to construct prompts as a causal intervention toward our SCM. Using graph sampling techniques and symbolic program executors, our strategy formalizes the structured causal prompts from commonsense knowledge bases. CLAP obtains state-of-the-art performance on WikiHow and RobotHow, achieving a relative improvement of 5.28% in human evaluations under the counterfactual setting. This indicates the superiority of CLAP in causal language planning semantically and sequentially.Preprint. Under review.

show abstract

Machine Translation to Sign Language Using Post-Translation Replacement Without Placeholders

Miyazaki¹,

Nakatani²,

Uchida³

et al. 2023

2023 IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops (ICASSPW)

View full text Add to dashboard Cite

GlobalWoZ: Globalizing MultiWoZ to Develop Multilingual Task-Oriented Dialogue Systems

Cited by 18 publications

References 20 publications

Multi2WOZ: A Robust Multilingual Dataset and Conversational Pretraining for Task-Oriented Dialog

Multi2WOZ: A Robust Multilingual Dataset and Conversational Pretraining for Task-Oriented Dialog

Neuro-Symbolic Causal Language Planning with Commonsense Prompting

Machine Translation to Sign Language Using Post-Translation Replacement Without Placeholders

Contact Info

Product

Resources

About