MDKB-Bot: A Practical Framework for Multi-Domain Task-Oriented Dialogue System

Four tables and two figures of this paper. Table 1 shows the statistics of training corpus for domain classification. Table 2 summarizes the statistics of training corpus for slot filling. Table 3 is a comparison of labeling performance on NLU. Table 4 shows dialogue quality evaluation results between top 4 teams in the competition. Figure 1 depicts the overall framework of the model consists of three components: (1) NLU module, which predicts intent domain and gives slot value of user utterance, (2) DM module, which outputs the dialogue action to NLG module, and (3) NLG module, which generates the final response. Figure 2 presents BLSTM-CRF model for slot filling.

本文共包含四张表格和两张图表。表格一展示了领域分类训练语料库的统计数据。表格二总结了领域分类训练语料库的槽位填充统计数据。表格三对比了自然语言理解（NLU）的标注性能。表格四展示了竞赛中排名前四的队伍之间的对话质量评估结果。图表一描绘了模型的整体框架，该框架由三个部分组成：（1）自然语言理解（NLU）模块，用于预测意图领域并给出用户语句的槽位值，（2）对话管理（DM）模块，向自然语言生成（NLG）模块输出对话动作，（3）自然语言生成（NLG）模块，生成最终的响应。图表二展示了用于槽位填充的BLSTM-CRF模型。