Dialogue State Tracking with a Language Model using Schema-Driven Prompting

EMNLP 2021  ·  Chia-Hsuan Lee, Hao Cheng, Mari Ostendorf ·

Task-oriented conversational systems often use dialogue state tracking to represent the user's intentions, which involves filling in values of pre-defined slots. Many approaches have been proposed, often using task-specific architectures with special-purpose classifiers. Recently, good results have been obtained using more general architectures based on pretrained language models. Here, we introduce a new variation of the language modeling approach that uses schema-driven prompting to provide task-aware history encoding that is used for both categorical and non-categorical slots. We further improve performance by augmenting the prompting with schema descriptions, a naturally occurring source of in-domain knowledge. Our purely generative system achieves state-of-the-art performance on MultiWOZ 2.2 and achieves competitive performance on two other benchmarks: MultiWOZ 2.1 and M2M. The data and code will be available at https://github.com/chiahsuan156/DST-as-Prompting.

PDF Abstract EMNLP 2021 PDF EMNLP 2021 Abstract

Datasets


Results from the Paper


 Ranked #1 on Dialogue State Tracking on MULTIWOZ 2.1 (MultiWOZ (Joint Goal Acc) metric)

     Get a GitHub badge
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Multi-domain Dialogue State Tracking MULTIWOZ 2.1 SGP-DST Joint Acc 56.66 # 7
Dialogue State Tracking MULTIWOZ 2.1 SGP-DST (base) MultiWOZ (Joint Goal Acc) 56.66 # 1
Dialogue State Tracking MULTIWOZ 2.1 SGP-DST (small) MultiWOZ (Joint Goal Acc) 56.12 # 2
Multi-domain Dialogue State Tracking MULTIWOZ 2.2 SGP-DST Joint Acc 57.6 # 3
Dialogue State Tracking MULTIWOZ 2.2 SGP-DST (base) MultiWOZ (Joint Goal Acc) 57.6 # 1
Dialogue State Tracking MULTIWOZ 2.2 SGP-DST (small) MultiWOZ (Joint Goal Acc) 56.3 # 2

Methods


No methods listed for this paper. Add relevant methods here