A.2 Datasets Table 5 shows the statistics of datasets, including the number of data and the average length of sentences. Similarly, "Tis" is topical word in the Shakespeareanplays. We compare baseline and many variant models of MPDL on TCFC dataset, the results are in Table 8. The supervised pipelined method, where the first model is to generate the response in styleS0 andthesecond model istotransfer itintotheresponse instyleS1 inasupervised manner (Pipeline). The non-parallel text transfer resources are easy to obtain.

Stylized dialogue generation, which aims to generate a given-style response for an input context, plays a vital role in intelligent dialogue systems.

Generative models have many applications in machine learning. To list a few, they have been usedtogenerate high-fidelity images [26,6],synthesize realistic speech andmusic fragments [58], improve the performance of semi-supervised learning [28, 10], detect adversarial examples and other anomalous data [54], imitation learning [22], and explore promising states in reinforcement learning [41].

However, these works face some limitations and challenges.

Layer normalization (LayerNorm) is a technique to normalize the distributions of intermediate layers. It enables smoother gradients, faster training, and better generalizationaccuracy.

Neural architecture search (NAS) is a popular area of machine learning, which aims to automate the process of developing neural architectures for a given dataset. Since 2017, a wide variety of NAS techniques have been proposed [78, 45, 32, 49].