MultiWOZ 2.2 : A Dialogue Dataset with Additional Annotation Corrections and State Tracking Baselines
By Xiaoxue Zang, Abhinav Rastogi, Jianguo Zhang, Jindong Chen. 2020
338 citations recorded by the source corpus.
MultiWOZ is a well-known task-oriented dialogue dataset containing over 10,000 annotated dialogues spanning 8 domains. It is extensively used as a benchmark for dialogue state tracking. However, recent works have reported presence of substantial noise in the dialogue state annotations. MultiWOZ 2.1 identified and fixed many of these erroneous annotations and user utterances, resulting in an improved version of this dataset. This work introduces MultiWOZ 2.2, which is a yet another improved version of this dataset. Firstly, we identify and fix dialogue state annotation errors across 17.3% of the utterances on top of MultiWOZ 2.1. Secondly, we redefine the ontology by disallowing vocabularies of slots with a large number of possible values (e.g., restaurant name, time of booking). In addition, we introduce slot span annotations for these slots to standardize them across recent models, which previously used custom string matching heuristics to generate them. We also benchmark a few state of the art dialogue state tracking models on the corrected dataset to facilitate comparison for future work. In the end, we discuss best practices for dialogue data collection that can help avoid annotation errors.
Subfields: Artificial Intelligence, Evaluation & Benchmarks, Dialogue & Conversational AI
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
- Jianfeng Gao
- Mohit Bansal
- Andrea Madotto
- Luke S. Zettlemoyer
- Percy Liang
- Yejin Choi
- Graham Neubig
- Richard Socher
In Evaluation & Benchmarks
In Dialogue & Conversational AI
- Jianfeng Gao
- Michel Galley
- Zhaojiang Lin
- Mari Ostendorf
- Zihan Liu
- Zhou Yu
- Chia-Hsuan Lee
- Andrea Madotto
Citations in the corpus
- Provably Confidential Language Modelling (2022, background): “MultiWOZ 2.2 is an already-public dialogue dataset written by crowd-workers, which collects over 10,000 annotated dialogues spanning 8 domains (Zang et al., 2020).”
- Language Models as Few-Shot Learner for Task-Oriented Dialogue Systems (2020, methodology): “ACT We use the MultiWoZ dataset for evaluating the speech ACT identification task.”
- Dialogue State Tracking with a Language Model using Schema-Driven Prompting (2021, methodology): “Since its initial release, many erroneous annotations and user utterances have been identified and fixed in subsequent versions, i.e., MultiWOZ 2.1 (Eric et al., 2019) and MultiWOZ 2.2 (Zang et al., 2020).”
- A Multitask, Multilingual, Multimodal Evaluation of ChatGPT on Reasoning, Hallucination, and Interactivity (2023, neutral): “2 (Zang et al., 2020).”
- GODEL: Large-Scale Pre-Training for Goal-Directed Dialog (2022, background): “Plato-XP does however provide some extrinsic evaluation on DSTC9-Track1 (Kim et al., 2020), MultiWOZ 2.2 (Zang et al., 2020), and DuConv (Wu et al., 2019), but is based on taskspecific automated metrics (ROUGE-L for DSTC9, Goal-Accuracy for MultiWOZ, and KnowledgeF1 for DuConv).”
- DeepStruct: Pretraining of Language Models for Structure Prediction (2022, methodology): “Datasets We use the MultiWOZ 2.1 (Budzianowski et al., 2018; Ramadan et al., 2018; Eric et al., 2020; Zang et al., 2020), which is a daily dialogue dataset for task-oriented conversations.”
- Leveraging Slot Descriptions for Zero-Shot Cross-Domain Dialogue StateTracking (2021, background): “Following recent works (Zhang et al., 2019; Rastogi et al., 2020), slots are divided into 2We are aware of STARC (Gao et al., 2020).”
- CUEing User Simulators: Calibrated User Embeddings for Multi-Turn Benchmarking (2026, unknown): “…(Li et al., 2018) 9,985 MuDoCo (Martin et al., 2020) 6,032 690 MultiDoGO (Peskov et al., 2019) 10,000 1,150 MultiWOZ 2.1 (Eric et al., 2020) 8,434 999 MultiWOZ 2.2 (Zang et al., 2020) 8,437 1,000 SGD (Rastogi et al., 2020) 10,000 2,482 STAR (Mosig et al., 2020) 6,145 • Distinctive Human Contrast.”