Sentiment-controlled text generation aims to generate texts according to the given sentiment. However, most of the existing studies focus only on the document- or sentence-level sentiment control, leaving a gap for finer-grained control over the content of generated results. Fine-grained control allows a generated review to express different opinions toward multiple aspects. Some previous works attempted to generate reviews conditioned on aspect-level sentiments, but they usually suffer from low adaptability and the lack of an annotated dataset. To alleviate these problems, we propose a novel pre-trained extended generative model that can dynamically refer to the prompt sentiment, together with an auxiliary classifier that extracts the fine-grained sentiments from the unannotated sentences, thus we conducted training on both annotated and unannotated datasets. We also propose a query-hint mechanism to further guide the generation process toward the aspect-level sentiments at every time step. Experimental results from real-world datasets demonstrated that our model has excellent adaptability in generating aspect-level sentiment-controllable review texts with high sentiment coverage and stable quality since, on both datasets, our model steadily outperforms other baseline models in the metrics of BLEU-4, METETOR, and ROUGE-L etc. The limitation of this work is that we only focus on fine-grained sentiments that are explicitly expressed. Moreover, the implicitly expressed fine-grained sentiment-controllable text generation will be an important puzzle for future work.

At present, sequence-based models have various applications in recommendation systems; these models recommend the interested items of the user according to the user’s behavioral sequence. However, sequence-based models have a limitation of length. When the length of the user’s behavioral sequence exceeds the limitation of the model, the model cannot take advantage of the complete behavioral sequence of the user and cannot know the user’s holistic interests. The accuracy of the model then goes down. Meanwhile, sequence-based models only pay attention to the sequential signals of the data but do not pay attention to the spatial signals of the data, which will also affect the model’s accuracy. This paper proposes a graph sequence-based model called GSRec that combines Graph Convolutional Network (GCN) and Transformer to solve these problems. In the GCN part we designed a reverse-order graph, and in the Transformer part we introduced the user embedding. The reverse-order graph and the user embedding can make the combination of GCN and Transformer more efficient. Experiments on six datasets show that GSRec outperforms the current state-of-the-art (SOTA) models.