Discovering Customer-Service Dialog System with Semi-Supervised Learning and Coarse-to-Fine Intent Detection
This work addresses data efficiency for customer-service dialog systems, but it is incremental as it builds on existing pre-trained models and modular approaches.
The paper tackled the problem of labeled-data scarcity in task-oriented dialog systems by constructing a weakly supervised dataset using a teacher/student paradigm and integrating coarse-to-fine intent detection, resulting in a higher success rate and more coherent responses.
Task-oriented dialog(TOD) aims to assist users in achieving specific goals through multi-turn conversation. Recently, good results have been obtained based on large pre-trained models. However, the labeled-data scarcity hinders the efficient development of TOD systems at scale. In this work, we constructed a weakly supervised dataset based on a teacher/student paradigm that leverages a large collection of unlabelled dialogues. Furthermore, we built a modular dialogue system and integrated coarse-to-fine grained classification for user intent detection. Experiments show that our method can reach the dialog goal with a higher success rate and generate more coherent responses.