近日，信息检索国际顶级会议(SIGIR 2024)公布了录用论文列表，博士生张晓堃关于推荐系统的两篇论文均被录取为长文。SIGIR是信息检索与数据挖掘领域的顶级会议，被CCF推荐为A类国际学术会议，在学术界及业界均享有盛誉。 这次会议共收到1148篇长文投稿，其中有791篇有效长文投稿，仅有159篇长文被录用，录用率仅为20%。

录用论文一: Disentangling ID and Modality Effects for Session-based Recommendation

Abstract: Session-based recommendation aims to predict intents of anonymous users based on their limited behaviors. Modeling user behaviors involves two distinct rationales: co-occurrence patterns reflected by item IDs, and fine-grained preferences represented by item modalities (text and images). However, existing methods typically entangle these causes, leading to their failure in understanding underlying reasons behind user interactions. Consequently, their recommendations suffer from inferior accuracy and lack of interpretability. To this end, we propose a novel framework called DIMO that effectively disentangles the effects of ID and modality, thereby improving both accuracy and interpretability of session-based recommendation. DIMO aims to disentangle these causes at both item and session levels. At the item level, we introduce a co-occurrence representation schema designed to explicitly incorporate co-occurrence patterns into ID representations. Simultaneously, DIMO aligns different modalities into a unified semantic space to represent them uniformly. As to the session level, we present a multi-view self-supervised disentanglement, including proxy mechanism and counterfactual inference, to disentangle ID and modality effects in the absence of supervised signals. Leveraging these disentangled causes, DIMO provides recommendations via causal inference and further creates two templates for generating explanations. Extensive experiments on multiple real-world datasets demonstrate the consistent superiority of DIMO over existing state-of-the-art methods, yielding significant improvements of up to 55% on common evaluation metrics. Further analysis also confirms DIMO's effectiveness in generating meaningful explanations.

中文摘要: 会话推荐旨在根据匿名用户有限的行为来预测其将来的行为。现有的推荐系统通常以两个不同的逻辑来揭露用户的行为模式：由商品ID反映的商品共现模式，以及由商品多模态信息(例如文本和图片)体现的用户细粒度偏好。然而，现有方法通常混淆了这两种基本的推荐逻辑，导致它们无法理解用户行为背后的根本原因。因此，它们的推荐精度较低，并缺乏可解释性。为此，本文提出了一种名为DIMO 的新型框架，该框架可以有效地对ID和模态所代表的不同逻辑进行解耦，从而提高会话推荐的准确度及可解释性。DIMO 旨在从商品和会话两个级别对ID和模态进行解耦。在商品级别，DIMO 提出了一个共现表示机制，旨在显式地将共现模式注入到ID的表示学习中。同时，DIMO 将异质的多模态模态信息对齐到一个统一的语义空间中，以对其进行统一的表示。在会话级别，DIMO提出了一个多视角自监督解耦算法，包括代理机制和反事实推断，以在没有监督信号的情况下对ID和模态信息进行解耦。利用解耦的ID和模态信息，DIMO 通过因果推断产生个性化推荐列表，并进一步创建了两个模板来对推荐结果生成解释说明。在多个真实数据集上进行的大量实验表明，DIMO 优于现有的方法，并产生了高达55%的显著性能提升。进一步的分析也证实了DIMO 可以对其产生的推荐结果生成有效的解释。

录用论文二: FineRec: Exploring Fine-grained Sequential Recommendation

Abstract: Sequential recommendation is dedicated to offering items of interest for users based on their history behaviors. The attribute-opinion pairs, expressed by users in their reviews for items, provide the potential to capture user preferences and item characteristics at a fine-grained level. To this end, we propose a novel framework FineRec that explores the attribute-opinion pairs of reviews to finely handle sequential recommendation. Specifically, we utilize a large language model to extract attribute-opinion pairs from the reviews. For each attribute, a unique attribute-specific user-opinion-item graph is created, where corresponding opinions serve as the edges linking heterogeneous user and item nodes. To tackle the diversity of opinions, we devise a diversity-aware convolution operation to aggregate information within the graphs, enabling attribute-specific user and item representation learning. Ultimately, we present an interaction driven fusion mechanism to integrate attribute-specific user/item representations across all attributes for generating final recommendations. Extensive experiments conducted on several real-world datasets demonstrate the superiority of the proposed FineRec over existing state-of-the-art methods. Further analysis also verifies the effectiveness of our fine-grained manner in handling the task.

中文摘要:序列推荐旨在根据用户的历史行为为其推荐感兴趣的商品。用户对商品评论中的属性-观点对以细粒度地方式体现了用户偏好和商品特征。为此，本文提出了一种新的框架 FineRec，通过探索评论中的属性-观点对来细粒度地处理序列推荐任务。具体来说，FineRec 利用大语言模型从评论中提取属性-观点对。对于每个属性，FineRec创建了一个属性特定的用户-观点-商品图，在这个图中，用户和商品构成了图的节点，而对应的观点作为连接用户和商品节点的边。为了处理观点的多样性，FineRec 设计了一个多样性感知的卷积操作来在图内聚合信息，实现特定属性下用户和商品的表示学习。然后，FineRec 提出了一个交互驱动的融合机制，以用户与商品的交互信息为信号将所有属性下的用户/商品表示进行融合，并以此生成最终的推荐列表。多个真实数据集上进行的大量实验表明，提出的 FineRec 显著优于目前最佳的序列推荐算法。进一步的分析也验证了本文提出的细粒度方式处理序列推荐任务的有效性。