OPERA: Harmonizing Task-Oriented Dialogs and Information Seeking Experience

Existing studies in conversational AI mostly treat task-oriented dialog (TOD) and question answering (QA) as separate tasks. Towards the goal of constructing a agent that can complete user tasks support information seeking, it is important to develop system handle both TOD QA with access various external knowledge sources. In this work, we propose new task, Open-Book (OB-TOD), which combines expands sources include explicit (e.g., web) implicit pre-trained language models). We create dataset OB-MultiWOZ, where enrich sessions QA-like information-seeking experience grounded on knowledge. unified model OPERA ( Op en-book E nd-to-end Task-o r iented Di log) appropriately tackle OB-TOD task. Experimental results show outperforms closed-book baselines, highlighting value types