国际万维网大会（The Web Conference，简称WWW会议）是由国际万维网会议委员会发起主办的国际顶级学术会议，创办于1994年，每年举办一届，是CCF-A类会议。WWW 2020将于2020年4月20日至4月24日在中国台湾台北举行。本届会议共收到了1129篇长文投稿，录用217篇长文，录用率为19.2%。
哈尔滨工业大学社会计算与信息检索研究中心有1篇长文被WWW 2020录用，下面是论文简要信息及摘要：
论文名称：Keywords Generation Improves E-Commerce Session-based Recommendation
作者：刘元兴，任昭春，张伟男，车万翔，刘挺，殷大伟
单位：哈尔滨工业大学，山东大学，京东
摘要：通过探索细粒度的用户行为，基于会话的推荐利用用户在短期内的行为预测用户的下一个动作。前人的工作仅仅利用了最后一次点击动作作为监督信号。在电商场景中，由于低包容性问题（即许多满足用户购物意图的相关产品被推荐系统所忽略），具有难以捉摸的点击行为和大规模的商品使这个任务具有挑战性。由于具有不同ID的相似产品可能具有相同的意图，因此我们认为，会话中的文本信息（例如，商品标题的关键字）可以用作额外的监督信号，以通过学习相似产品中更多的共同意图来解决上述问题。因此，为了提高基于电商会话的推荐的性能，我们根据当前会话中的点击顺序生成关键字来推断用户的意图。
在本文中，我们提出了带有关键字生成的基于电商会话的推荐模型（ESRM-KG）。具体地，ESRM-KG模型首先将输入的点击序列编码为高维向量表示；然后利用一种双线性解码，预测当前会话中的下一个动作；同时ESRM-KG模型处理其编码器的高维表示，以为整个会话生成可解释的关键字。我们在大规模的电商数据集上进行了大量的实验。我们的实验结果表明，借助关键字生成，ESRM-KG模型的性能优于最新的基线。我们还通过样例分析来讨论关键字生成如何帮助基于电商会话的推荐。
Abstract: By exploring fine-grained user behaviors, session-based recommendation predicts a user’s next action from short-term behavior sessions. Most of the previous work learns about a user’s implicit behavior by merely taking the last click action as the supervision signal. However, in e-commerce scenarios, large-scale products with elusive click behaviors make such task challenging because of the low inclusiveness problem, i.e., many relevant products that satisfy the user’s shopping intention are neglected by recommenders. Since similar products with different IDs may share the same intention, we argue that the textual information (e.g., keywords of product titles) from sessions can be used as additional supervision signals to tackle the above problem through learning more shared intention within similar products. Therefore, to improve the performance of e-commerce session-based recommendation, we explicitly infer the user’s intention by generating keywords entirely from the click sequence in the current session.
In this paper, we propose the e-commerce session-based recommendation model with keywords generation (abbreviated as ESRM-KG) to integrate keywords generation into e-commerce session-based recommendation. Specifically, the ESRM-KG model firstly encodes an input action sequence into a high dimensional representation; then it presents a bi-linear decoding scheme to predict the next action in the current session; synchronously, the ESRM-KG model addresses incepts the high dimensional representation of its encoder to generate explainable keywords for the whole session. We carried out extensive experiments in the context of click prediction on a large-scale real-world e-commerce dataset. Our experimental results show that the ESRM-KGmodel outperforms state-of-the-art baselines with the help of keywords generation. We also discuss how keywords generation helps the e-commerce session-based recommendation with case studies and error analysis.
