Schema-Guided Semantic Accuracy: Faithfulness in Task-Oriented Dialogue Response Generation
This work addresses the problem of improving evaluation and generation faithfulness in task-oriented dialogue systems, which is incremental as it builds on existing methods like T5 and prefix tuning.
The paper tackled the problem of evaluating faithfulness in task-oriented dialogue response generation by proposing Schema-Guided Semantic Accuracy (SGSAcc) to assess utterances from both categorical and non-categorical slots, showing good agreement with human judgment on the SGD dataset. It also identified a weakness in generating faithful utterances for categorical slots in unseen domains and addressed it with prefix tuning on T5, achieving the lowest Slot Error Rate (SER) reported and high SGSAcc.
Ensuring that generated utterances are faithful to dialogue actions is crucial for Task-Oriented Dialogue Response Generation. Slot Error Rate (SER) only partially measures generation quality in that it solely assesses utterances generated from non-categorical slots whose values are expected to be reproduced exactly. Utterances generated from categorical slots, which are more variable, are not assessed by SER. We propose Schema-Guided Semantic Accuracy (SGSAcc) to evaluate utterances generated from both categorical and non-categorical slots by recognizing textual entailment. We show that SGSAcc can be applied to evaluate utterances generated from a wide range of dialogue actions in the Schema Guided Dialogue (SGD) dataset with good agreement with human judgment. We also identify a previously overlooked weakness in generating faithful utterances from categorical slots in unseen domains. We show that prefix tuning applied to T5 generation can address this problem. We further build an ensemble of prefix-tuning and fine-tuning models that achieves the lowest SER reported and high SGSAcc on the SGD dataset.