تم تطبيق نماذج التسلسل إلى التسلسل على مجموعة واسعة من مهام NLP، ولكن كيفية استخدامها بشكل صحيح لتتبع حالة الحوار بشكل منهجي. في هذه الورقة، ندرس هذه المشكلة من وجهات نظر أهداف ما قبل التدريب وكذلك تنسيقات تمثيلات السياق. نوضح أن اختيار الهدف ما قبل التدريب يجعل فرقا كبيرا لجودة تتبع الدولة. على وجه الخصوص، نجد أن التنبؤ الأمان المقنع هو أكثر فعالية من نمذجة اللغة التراجع التلقائي. نستكشف أيضا استخدام Pegasus، وهو هدف ما قبل التدريب المستندة إلى التنبؤ بتلخيص النص، لنموذج تتبع الدولة. وجدنا أن التدريب المسبق لمهمة التلخيص البعيدة على ما يبدو يعمل بشكل جيد بشكل جيد لتتبع حالة الحوار. بالإضافة إلى ذلك، وجدنا أنه في حين أن تمثيل سياق الدولة المتكرر يعمل أيضا بشكل جيد بشكل معقول، فقد يكون للنموذج صعوبة في التعافي من الأخطاء السابقة. أجرينا تجارب في مجموعات بيانات MultiWoz 2.1-2.4 و Woz 2.0 و DSTC2 مع ملاحظات متسقة.
Sequence-to-sequence models have been applied to a wide variety of NLP tasks, but how to properly use them for dialogue state tracking has not been systematically investigated. In this paper, we study this problem from the perspectives of pre-training objectives as well as the formats of context representations. We demonstrate that the choice of pre-training objective makes a significant difference to the state tracking quality. In particular, we find that masked span prediction is more effective than auto-regressive language modeling. We also explore using Pegasus, a span prediction-based pre-training objective for text summarization, for the state tracking model. We found that pre-training for the seemingly distant summarization task works surprisingly well for dialogue state tracking. In addition, we found that while recurrent state context representation works also reasonably well, the model may have a hard time recovering from earlier mistakes. We conducted experiments on the MultiWOZ 2.1-2.4, WOZ 2.0, and DSTC2 datasets with consistent observations.
References used
https://aclanthology.org/
Abstract Tracking dialogue states to better interpret user goals and feed downstream policy learning is a bottleneck in dialogue management. Common practice has been to treat it as a problem of classifying dialogue content into a set of pre-defined s
Zero-shot transfer learning for dialogue state tracking (DST) enables us to handle a variety of task-oriented dialogue domains without the expense of collecting in-domain data. In this work, we propose to transfer the cross-task knowledge from genera
Dialogue state tracking models play an important role in a task-oriented dialogue system. However, most of them model the slot types conditionally independently given the input. We discover that it may cause the model to be confused by slot types tha
This paper aims at providing a comprehensive overview of recent developments in dialogue state tracking (DST) for task-oriented conversational systems. We introduce the task, the main datasets that have been exploited as well as their evaluation metr
Recently, the focus of dialogue state tracking has expanded from single domain to multiple domains. The task is characterized by the shared slots between domains. As the scenario gets more complex, the out-of-vocabulary problem also becomes severer.