Inductive knowledge graph embedding (KGE) aims to embed unseen entities in emerging knowledge graphs (KGs). The major recent studies of inductive KGE embed unseen entities by aggregating information from their neighboring entities and relations with graph neural networks (GNNs). However, these methods rely on the existing neighbors of unseen entities and suffer from two common problems: data sparsity and feature smoothing. Firstly, the data sparsity problem means unseen entities usually emerge with few triplets containing insufficient information. Secondly, the effectiveness of the features extracted from original KGs will degrade when repeatedly propagating these features to represent unseen entities in emerging KGs, which is termed feature smoothing problem. To tackle the two problems, we propose a novel model entitled Meta-Learning Based Memory Graph Convolutional Network (MMGCN) consisting of three different components: 1) the two-layer information transforming module (TITM) developed to effectively transform information from original KGs to emerging KGs; 2) the hyper-relation feature initializing module (HFIM) proposed to extract type-level features shared between KGs and obtain a coarse-grained representation for each entity with these features; and 3) the meta-learning training module (MTM) designed to simulate the few-shot emerging KGs and train the model in a meta-learning framework. The extensive experiments conducted on the few-shot link prediction task for emerging KGs demonstrate the superiority of our proposed model MMGCN compared with state-of-the-art methods.

Author name disambiguation (AND) is a central task in academic search, which has received more attention recently accompanied by the increase of authors and academic publications. To tackle the AND problem, existing studies have proposed various approaches based on different types of information, such as raw document features (e.g., co-authors, titles, and keywords), the fusion feature (e.g., a hybrid publication embedding based on multiple raw document features), the local structural information (e.g., a publication's neighborhood information on a graph), and the global structural information (e.g., interactive information between a node and others on a graph). However, there has been no work taking all the above-mentioned information into account and taking full advantage of the contributions of each raw document feature for the AND problem so far. To fill the gap, we propose a novel framework named EAND (Towards Effective Author Name Disambiguation by Hybrid Attention). Specifically, we design a novel feature extraction model, which consists of three hybrid attention mechanism layers, to extract key information from the global structural information and the local structural information that are generated from six similarity graphs constructed based on different similarity coefficients, raw document features, and the fusion feature. Each hybrid attention mechanism layer contains three key modules: a local structural perception, a global structural perception, and a feature extractor. Additionally, the mean absolute error function in the joint loss function is used to introduce the structural information loss of the vector space. Experimental results on two real-world datasets demonstrate that EAND achieves superior performance, outperforming state-of-the-art methods by at least +2.74% in terms of the micro-F1 score and +3.31% in terms of the macro-F1 score.

Entity linking is a new technique in recommender systems to link users’ interaction behaviors in different domains, for the purpose of improving the performance of the recommendation task. Linking-based cross-domain recommendation aims to alleviate the data sparse problem by utilizing the domain-sharable knowledge from auxiliary domains. However, existing methods fail to prevent domain-specific features to be transferred, resulting in suboptimal results. In this paper, we aim to address this issue by proposing an adversarial transfer learning based model ATLRec, which effectively captures domain-sharable features for cross-domain recommendation. In ATLRec, we leverage adversarial learning to generate representations of user-item interactions in both the source and the target domains, such that the discriminator cannot identify which domain they belong to, for the purpose of obtaining domain-sharable features. Meanwhile each domain learns its domain-specific features by a private feature extractor. The recommendation of each domain considers both domain-specific and domain-sharable features. We further adopt an attention mechanism to learn item latent factors of both domains by utilizing the shared users with interaction history, so that the representations of all items can be learned sufficiently in a shared space, even when few or even no items are shared by different domains. By this method, we can represent all items from the source and the target domains in a shared space, for the purpose of better linking items in different domains and capturing cross-domain item-item relatedness to facilitate the learning of domain-sharable knowledge. The proposed model is evaluated on various real-world datasets and demonstrated to outperform several state-of-the-art single-domain and cross-domain recommendation methods in terms of recommendation accuracy.

Entity linking (EL) is the task of determining the identity of textual entity mentions given a predefined knowledge base (KB). Plenty of existing efforts have been made on this task using either “local” information (contextual information of the mention in the text), or “global” information (relations among candidate entities). However, either local or global information might be insufficient especially when the given text is short. To get richer local and global information for entity linking, we propose to enrich the context information for mentions by getting extra contexts from the web through web search engines (WSE). Based on the intuition above, two novel attempts are made. The first one adds web-searched results into an embedding-based method to expand the mention’s local information, where we try two different methods to help generate high-quality web contexts: one is to apply the attention mechanism and the other is to use the abstract extraction method. The second one uses the web contexts to extend the global information, i.e., finding and utilizing more extra relevant mentions from the web contexts with a graph-based model. Finally, we combine the two models we propose to use both extended local and global information from the extra web contexts. Our empirical study based on six real-world datasets shows that using extra web contexts to extend the local and the global information could effectively improve the F1 score of entity linking.

As a fundamental operation in LBS (location-based services), the trajectory similarity of moving objects has been extensively studied in recent years. However, due to the increasing volume of moving object trajectories and the demand of interactive query performance, the trajectory similarity queries are now required to be processed on massive datasets in a real-time manner. Existing work has proposed distributed or parallel solutions to enable large-scale trajectory similarity processing. However, those techniques cannot be directly adapted to the real-time scenario as it is likely to generate poor balancing performance when workload variance occurs on the incoming trajectory stream. In this paper, we propose a new workload partitioning framework, ART (Adaptive Framework for Real-Time Trajectory Similarity), which introduces practical algorithms to support dynamic workload assignment for RTTS (real-time trajectory similarity). Our proposal includes a processing model tailored for the RTTS scenario, a load balancing framework to maximize throughput, and an adaptive data partition manner designed to cut off unnecessary network cost. Based on this, our model can handle the large-scale trajectory similarity in an on-line scenario, which achieves scalability, effectiveness, and efficiency by a single shot. Empirical studies on synthetic data and real-world stream applications validate the usefulness of our proposal and prove the huge advantage of our approach over state-of-the-art solutions in the literature.

With the development and prevalence of online social networks, there is an obvious tendency that people are willing to attend and share group activities with friends or acquaintances. This motivates the study on group recommendation, which aims to meet the needs of a group of users, instead of only individual users. However, how to aggregate different preferences of different group members is still a challenging problem: 1) the choice of a member in a group is influenced by various factors, e.g., personal preference, group topic, and social relationship; 2) users have different influences when in different groups. In this paper, we propose a generative geo-social group recommendation model (GSGR) to recommend points of interest (POIs) for groups. Specifically, GSGR well models the personal preference impacted by geographical information, group topics, and social influence for recommendation. Moreover, when making recommendations, GSGR aggregates the preferences of group members with different weights to estimate the preference score of a group to a POI. Experimental results on two datasets show that GSGR is effective in group recommendation and outperforms the state-of-the-art methods.