Augmenting Sequential Recommendation with Pseudo-Prior Items via Reversely Pre-training Transformer
This addresses the cold-start problem for users with limited interaction history in recommendation systems, representing an incremental improvement over existing transformer-based methods.
The paper tackles the cold-start issue in sequential recommendation, where transformer-based models perform poorly for short sequences, by proposing ASReP, a framework that augments short sequences with pseudo-prior items generated via reverse pre-training, achieving improved performance on two real-world datasets.
Sequential Recommendation characterizes the evolving patterns by modeling item sequences chronologically. The essential target of it is to capture the item transition correlations. The recent developments of transformer inspire the community to design effective sequence encoders, \textit{e.g.,} SASRec and BERT4Rec. However, we observe that these transformer-based models suffer from the cold-start issue, \textit{i.e.,} performing poorly for short sequences. Therefore, we propose to augment short sequences while still preserving original sequential correlations. We introduce a new framework for \textbf{A}ugmenting \textbf{S}equential \textbf{Re}commendation with \textbf{P}seudo-prior items~(ASReP). We firstly pre-train a transformer with sequences in a reverse direction to predict prior items. Then, we use this transformer to generate fabricated historical items at the beginning of short sequences. Finally, we fine-tune the transformer using these augmented sequences from the time order to predict the next item. Experiments on two real-world datasets verify the effectiveness of ASReP. The code is available on \url{https://github.com/DyGRec/ASReP}.