Alternating Recurrent Dialog Model with Large-scale Pre-trained Language Models

Existing dialog system models require extensive human annotations and are difficult to generalize to different tasks. The recent success of large pre-trained language models has suggested the effectiveness of incorporating language priors in down-stream NLP tasks. However, how much pre-trained language models can help dialog response generation is still under exploration. In this paper, we propose a simple, general, and effective framework: Alternating Recurrent Dialog Model (ARDM). ARDM models each speaker separately and takes advantage of large pre-trained language models. It requires no supervision from human annotations such as belief states or dialog acts to achieve effective conversations. ARDM outperforms or is on par with the state-of-theart methods on two popular task-oriented dialog datasets: CamRest676 and MultiWOZ. Moreover, we can generalize ARDM to more challenging, non-collaborative tasks such as persuasion. In the PersuasionForGood task, ARDM is capable of generating human-like responses to persuade people to donate to a charity.1

Introduction. It has been a long-standing ambition for artificial intelligence researchers to create an intelligent conversational agent that can generate human-like responses. Recently, data-driven dialog models are more and more popular. However, most current state-of-the-art approaches still heavily rely on extensive human annotations such as belief states and dialog acts (Lei et al., 2018). However, dialog content can vary considerably in different dialog tasks. Having a different intent or dialog act annotation scheme for each task is costly and even impossible for tasks such as open-domain social chat. Thus, it is difficult to utilize these methods on challenging dialog tasks where dialog states and acts are difficult to annotate such as persuasion and negotiation. Eric and Manning (2017) proposed a simple sequence-to-sequence architecture that requires no explicit annotations. The model learns to extract information from dialog history with attention and copy mechanism.

Discussion / Conclusion. ARDM models speakers separately on top of a large pre-trained language model. Such simple adaptation demonstrates substantial performance gain. We suspect it is because the interleaved structure of two language models provides a collabora- We propose to build Alternating Recurrent Dialog Model (ARDM), a simple, general, and effective dialog method that models user and system separately with large-scale pre-trained language models. Since ARDM does not require any annotations, it generalizes to different dialog applications. Experimental results on CamRest676 and MultiWOZ suggest that ARDM outperforms or is on-par with the current state-of-the-art methods that use manual annotation information, such as belief states and dialog acts. Furthermore, we find our model’s excellent performance generalizes to more complex non-collaborative dialog settings. It can generate high-quality responses to persuade people to donate to charity. However, the easiness of training ARDM raises concerns about the misuse of the model in scenarios such as sales, harassment, or scam on a mass scale. We caution the public in deploying such systems in the real world.