Iterative rule-guided reasoning over sparse knowledge graphs with deep reinforcement learning

In recent years, reasoning over knowledge graphs (KGs) has been widely adapted to empower retrieval systems, recommender systems, and question answering systems, generating a surge in research interest. Recently developed reasoning methods usually suffer from poor performance when applied to incomplete or sparse KGs, due to the lack of evidential paths that can reach target entities. To solve this problem, we propose a hybrid multi-hop reasoning model with reinforcement learning (RL) called SparKGR, which implements dynamic path completion and iterative rule guidance strategies to increase reasoning performance over sparse KGs. Firstly, the model dynamically completes the missing paths using rule guidance to augment the action space for the RL agent; this strategy effectively reduces the sparsity of KGs, thus increasing path search efficiency. Secondly, an iterative optimization of rule induction and fact inference is designed to incorporate global information from KGs to guide the RL agent exploration; this optimization iteratively improves overall training performance. We further evaluated the SparKGR model through different tasks on five real world datasets extracted from Freebase, Wikidata and NELL. The experimental results indicate that SparKGR outperforms state-of-the-art baseline models without losing interpretability.     

Multi-heterogeneous neighborhood-aware for Knowledge Graphs alignment

Entity alignment is an important task for the Knowledge Graph (KG) completion, which aims to identify the same entities in different KGs. Most of previous works only utilize the relation structures of KGs, but ignore the heterogeneity of relations and attributes of KGs. However, these information can provide more feature information and improve the accuracy of entity alignment. In this paper, we propose a novel Multi-Heterogeneous Neighborhood-Aware model (MHNA) for KGs alignment. MHNA aggregates multi-heterogeneous information of aligned entities, including the entity name, relations, attributes and attribute values. An important contribution is to design a variant attention mechanism, which adds the feature information of relations and attributes to the calculation of attention coefficients. Extensive experiments on three well-known benchmark datasets show that MHNA significantly outperforms 12 state-of-the-art approaches, demonstrating that our approach has good scalability and superiority in both cross-language and monolingual KGs. An ablation study further supports the effectiveness of our variant attention mechanism.     

An efficiency relation-specific graph transformation network for knowledge graph representation learning

Knowledge graph representation learning (KGRL) aims to infer the missing links between target entities based on existing triples. Graph neural networks (GNNs) have been introduced recently as one of the latest trendy architectures serves KGRL task using aggregations of neighborhood information. However, current GNN-based methods have fundamental limitations in both modelling the multi-hop distant neighbors and selecting relation-specific neighborhood information from vast neighbors. In this study, we propose a new relation-specific graph transformation network (RGTN) for the KGRL task. Specifically, the proposed RGTN is the first pioneer model that transforms a relation-based graph into a new path-based graph by generating useful paths that connect heterogeneous relations and multi-hop neighbors. Unlike the existing GNN-based methods, our approach is able to adaptively select the most useful paths for each specific relation and to effectively build path-based connections between unconnected distant entities. The transformed new graph structure opens a new way to model the arbitrary lengths of multi-hop neighbors which leads to more effective embedding learning. In order to verify the effectiveness of our proposed model, we conduct extensive experiments on three standard benchmark datasets, e.g., WN18RR, FB15k-237 and YAGO-10-DR. Experimental results show that the proposed RGTN achieves the promising results and even outperforms other state-of-the-art models on the KGRL task (e.g., compared to other state-of-the-art GNN-based methods, our model achieves 2.5% improvement using H@10 on WN18RR, 1.2% improvement using H@10 on FB15k-237 and 6% improvement using H@10 on YAGO3-10-DR).     

GAFM: A Knowledge Graph Completion Method Based on Graph Attention Faded Mechanism

Although the Knowledge Graph (KG) has been successfully applied to various applications, there is still a large amount of incomplete knowledge in the KG. This study proposes a Knowledge Graph Completion (KGC) method based on the Graph Attention Faded Mechanism (GAFM) to solve the problem of incomplete knowledge in KG. GAFM introduces a graph attention network that incorporates the information in multi-hop neighborhood nodes to embed the target entities into low dimensional space. To generate a more expressive entity representation, GAFM gives different weights to the neighborhood nodes of the target entity by adjusting the attention value of neighborhood nodes according to the variation of the path length. The attention value is adjusted by the attention faded coefficient, which decreases with the increase of the distance between the neighborhood node and the target entity. Then, considering that the capsule network has the ability to fit features, GAFM introduces the capsule network as the decoder to extract feature information from triple representations. To verify the effectiveness of the proposed method, we conduct a series of comparative experiments on public datasets (WN18RR and FB15k-237). Experimental results show that the proposed method outperforms baseline methods. The Hits@10 metric is improved by 8% compared with the second-place KBGAT method.     

Cascade embedding model for knowledge graph inference and retrieval

Knowledge graphs are widely used in retrieval systems, question answering systems (QA), hypothesis generation systems, etc. Representation learning provides a way to mine knowledge graphs to detect missing relations; and translation-based embedding models are a popular form of representation model. Shortcomings of translation-based models however, limits their practicability as knowledge completion algorithms. The proposed model helps to address some of these shortcomings.The similarity between graph structural features of two entities was found to be correlated to the relations of those entities. This correlation can help to solve the problem caused by unbalanced relations and reciprocal relations. We used Node2vec, a graph embedding algorithm, to represent information related to an entity's graph structure, and we introduce a cascade model to incorporate graph embedding with knowledge embedding into a unified framework. The cascade model first refines feature representation in the first two stages (Local Optimization Stage), and then uses backward propagation to optimize parameters of all the stages (Global Optimization Stage). This helps to enhance the knowledge representation of existing translation-based algorithms by taking into account both semantic features and graph features and fusing them to extract more useful information. Besides, different cascade structures are designed to find the optimal solution to the problem of knowledge inference and retrieval.The proposed model was verified using three mainstream knowledge graphs: WIN18, FB15K and BioChem. Experimental results were validated using the hit@10 rate entity prediction task. The proposed model performed better than TransE, giving an average improvement of 2.7% on WN18, 2.3% on FB15k and 28% on BioChem. Improvements were particularly marked where there were problems with unbalanced relations and reciprocal relations. Furthermore, the stepwise-cascade structure is proved to be more effective and significantly outperforms other baselines.     

Aspect sentiment analysis with heterogeneous graph neural networks

Aspect-based sentiment analysis technologies may be a very practical methodology for securities trading, commodity sales, movie rating websites, etc. Most recent studies adopt the recurrent neural network or attention-based neural network methods to infer aspect sentiment using opinion context terms and sentence dependency trees. However, due to a sentence often having multiple aspects sentiment representation, these models are hard to achieve satisfactory classification results. In this paper, we discuss these problems by encoding sentence syntax tree, words relations and opinion dictionary information in a unified framework. We called this method heterogeneous graph neural networks (Hete_GNNs). Firstly, we adopt the interactive aspect words and contexts to encode the sentence sequence information for parameter sharing. Then, we utilized a novel heterogeneous graph neural network for encoding these sentences’ syntax dependency tree, prior sentiment dictionary, and some part-of-speech tagging information for sentiment prediction. We perform the Hete_GNNs sentiment judgment and report the experiments on five domain datasets, and the results confirm that the heterogeneous context information can be better captured with heterogeneous graph neural networks. The improvement of the proposed method is demonstrated by aspect sentiment classification task comparison.     

Boosting question answering over knowledge graph with reward integration and policy evaluation under weak supervision

Among existing knowledge graph based question answering (KGQA) methods, relation supervision methods require labeled intermediate relations for stepwise reasoning. To avoid this enormous cost of labeling on large-scale knowledge graphs, weak supervision methods, which use only the answer entity to evaluate rewards as supervision, have been introduced. However, lacking intermediate supervision raises the issue of sparse rewards, which may result in two types of incorrect reasoning path: (1) incorrectly reasoned relations, even when the final answer entity may be correct; (2) correctly reasoned relations in a wrong order, which leads to an incorrect answer entity. To address these issues, this paper considers the multi-hop KGQA task as a Markov decision process, and proposes a model based on Reward Integration and Policy Evaluation (RIPE). In this model, an integrated reward function is designed to evaluate the reasoning process by leveraging both terminal and instant rewards. The intermediate supervision for each single reasoning hop is constructed with regard to both the fitness of the taken action and the evaluation of the unreasoned information remained in the updated question embeddings. In addition, to lead the agent to the answer entity along the correct reasoning path, an evaluation network is designed to evaluate the taken action in each hop. Extensive ablation studies and comparative experiments are conducted on four KGQA benchmark datasets. The results demonstrate that the proposed model outperforms the state-of-the-art approaches in terms of answering accuracy.     

EGC: A novel event-oriented graph clustering framework for social media text

With the popularity of social platforms such as Sina Weibo, Tweet, etc., a large number of public events spread rapidly on social networks and huge amount of textual data are generated along with the discussion of netizens. Social text clustering has become one of the most critical methods to help people find relevant information and provides quality data for subsequent timely public opinion analysis. Most existing neural clustering methods rely on manual labeling of training sets and take a long time in the learning process. Due to the explosiveness and the large-scale of social media data, it is a challenge for social text data clustering to satisfy the timeliness demand of users. This paper proposes a novel unsupervised event-oriented graph clustering framework (EGC), which can achieve efficient clustering performance on large-scale datasets with less time overhead and does not require any labeled data. Specifically, EGC first mines the potential relations existing in social text data and transforms the textual data of social media into an event-oriented graph by taking advantage of graph structure for complex relations representation. Secondly, EGC uses a keyword-based local importance method to accurately measure the weights of relations in event-oriented graph. Finally, a bidirectional depth-first clustering algorithm based on the interrelations is proposed to cluster the nodes in event-oriented graph. By projecting the relations of the graph into a smaller domain, EGC achieves fast convergence. The experimental results show that the clustering performance of EGC on the Weibo dataset reaches 0.926 (NMI), 0.926 (AMI), 0.866 (ARI), which are 13%–30% higher than other clustering methods. In addition, the average query time of EGC clustered data is 16.7ms, which is 90% less than the original data.     

Global information-aware argument mining based on a top-down multi-turn QA model

Argument mining (AM) aims to automatically generate a graph that represents the argument structure of a document. Most previous AM models only pay attention to a single argument component (AC) to classify the type of the AC or a pair of ACs to identify and classify the argumentative relation (AR) between the two ACs. These models ignore the impact of global argument structure of the documents, which is important, especially in some highly structured genres such as scientific papers, where the process of argumentation is relatively fixed. Inspired by this, we propose a novel two-stage model which leverages global structure information to support AM. The first stage uses a multi-turn question-answering model to incrementally generate an initial argumentative graph that identifies relations among ACs. At each turn, all ACs related to the query AC are generated simultaneously, such that the sibling global information between the answer ACs is considered. In addition, the partially constructed graph is used as global structure information to support the extension of the graph with additional ACs. After the whole initial graph structure has been determined, the second stage assigns semantic types to both the ACs and ARs among them, leveraging information from this initial graph as global structure information. We test the proposed methods on two scientific datasets (one is the AbstRCT dataset including 659 abstracts about cancer research and the other is the SciARG dataset that consists of 225 computer linguistic abstracts and 285 biomedical abstracts) and a student essay dataset PE with 402 essays. Our experiments show that our model improves the state-of-the-art performance on two scientific datasets for different AM subtasks, with average improvements of 1%, 2.41%, 1.1% for the ACC, ARI and ARC task respectively on the AbstRCT dataset, and 2.36%, 1.84%, 8.87% for the ACC, ARI and ARC task on the SciARG dataset. Our model also achieves comparative results on the PE datasets: 87.7% of F1 scores for the ACC task, 81.4% for the ARI task and 78.8% for the ARC task.     

Adaptive dictionary and structure learning for unsupervised feature selection

Unsupervised feature selection is very attractive in many practical applications, as it needs no semantic labels during the learning process. However, the absence of semantic labels makes the unsupervised feature selection more challenging, as the method can be affected by the noise, redundancy, or missing in the originally extracted features. Currently, most methods either consider the influence of noise for sparse learning or think over the internal structure information of the data, leading to suboptimal results. To relieve these limitations and improve the effectiveness of unsupervised feature selection, we propose a novel method named Adaptive Dictionary and Structure Learning (ADSL) that conducts spectral learning and sparse dictionary learning in a unified framework. Specifically, we adaptively update the dictionary based on sparse dictionary learning. And, we also introduce the spectral learning method of adaptive updating affinity matrix. While removing redundant features, the intrinsic structure of the original data can be retained. In addition, we adopt matrix completion in our framework to make it competent for fixing the missing data problem. We validate the effectiveness of our method on several public datasets. Experimental results show that our model not only outperforms some state-of-the-art methods on complete datasets but also achieves satisfying results on incomplete datasets.     

Robust graph learning with graph convolutional network

Graph convolutional network (GCN) is a powerful tool to process the graph data and has achieved satisfactory performance in the task of node classification. In general, GCN uses a fixed graph to guide the graph convolutional operation. However, the fixed graph from the original feature space may contain noises or outliers, which may degrade the effectiveness of GCN. To address this issue, in this paper, we propose a robust graph learning convolutional network (RGLCN). Specifically, we design a robust graph learning model based on the sparse constraint and strong connectivity constraint to achieve the smoothness of the graph learning. In addition, we introduce graph learning model into GCN to explore the representative information, aiming to learning a high-quality graph for the downstream task. Experiments on citation network datasets show that the proposed RGLCN outperforms the existing comparison methods with respect to the task of node classification.     

Phishing scams detection via temporal graph attention network in Ethereum

Recently, phishing scams have become one of the most serious types of crime involved in Ethereum, the second-largest blockchain-based cryptocurrency platform. The existing phishing scams detection techniques for Ethereum mostly use traditional machine learning or network representation learning to mine the key information from the transaction network and identify phishing addresses. However, these methods typically crop the temporal transaction graph into snapshot sequences or construct temporal random wanderings to model the dynamic evolution of the topology of the transaction graph. In this paper, we propose PDTGA, a method that applies graph representation learning based on temporal graphs attention to improve the effectiveness of phishing scams detection in Ethereum. Specifically, we learn the functional representation of time directly and model the time signal through the interactions between the time encoding function and node features, edge features, and the topology of the graph. We collected a real-world Ethereum phishing scam dataset, containing over 250,000 transaction records between more than 100,000 account addresses, and divided them into three datasets of different sizes. Through data analysis, we first summarized the periodic pattern of Ethereum phishing scam activities. Then we constructed 14 kinds of account node features and 3 kinds of transaction edge features. Experimental evaluations based on the above three datasets demonstrate that PDTGA with 94.78% AUC score and 88.76% recall score outperforms the state-of-the-art methods.     

Contrastive Graph Convolutional Networks with adaptive augmentation for text classification

Text classification is an important research topic in natural language processing (NLP), and Graph Neural Networks (GNNs) have recently been applied in this task. However, in existing graph-based models, text graphs constructed by rules are not real graph data and introduce massive noise. More importantly, for fixed corpus-level graph structure, these models cannot sufficiently exploit the labeled and unlabeled information of nodes. Meanwhile, contrastive learning has been developed as an effective method in graph domain to fully utilize the information of nodes. Therefore, we propose a new graph-based model for text classification named CGA2TC, which introduces contrastive learning with an adaptive augmentation strategy into obtaining more robust node representation. First, we explore word co-occurrence and document word relationships to construct a text graph. Then, we design an adaptive augmentation strategy for the text graph with noise to generate two contrastive views that effectively solve the noise problem and preserve essential structure. Specifically, we design noise-based and centrality-based augmentation strategies on the topological structure of text graph to disturb the unimportant connections and thus highlight the relatively important edges. As for the labeled nodes, we take the nodes with same label as multiple positive samples and assign them to anchor node, while we employ consistency training on unlabeled nodes to constrain model predictions. Finally, to reduce the resource consumption of contrastive learning, we adopt a random sample method to select some nodes to calculate contrastive loss. The experimental results on several benchmark datasets can demonstrate the effectiveness of CGA2TC on the text classification task.     

Reinforcement learning with dynamic completion for answering multi-hop questions over incomplete knowledge graph

Text-enhanced and implicit reasoning methods are proposed for answering questions over incomplete knowledge graph (KG), whereas prior studies either rely on external resources or lack necessary interpretability. This article desires to extend the line of reinforcement learning (RL) methods for better interpretability and dynamically augment original KG action space with additional actions. To this end, we propose a RL framework along with a dynamic completion mechanism, namely Dynamic Completion Reasoning Network (DCRN). DCRN consists of an action space completion module and a policy network. The action space completion module exploits three sub-modules (relation selector, relation pruner and tail entity predictor) to enrich options for decision making. The policy network calculates probability distribution over joint action space and selects promising next-step actions. Simultaneously, we employ the beam search-based action selection strategy to alleviate delayed and sparse rewards. Extensive experiments conducted on WebQSP, CWQ and MetaQA demonstrate the effectiveness of DCRN. Specifically, under 50% KG setting, the Hits@1 performance improvements of DCRN on MetaQA-1H and MetaQA-3H are 2.94% and 1.18% respectively. Moreover, under 30% and 10% KG settings, DCRN prevails over all baselines by 0.9% and 1.5% on WebQSP, indicating the robustness to sparse KGs.     

Graph neural news recommendation with long-term and short-term interest modeling

With the information explosion of news articles, personalized news recommendation has become important for users to quickly find news that they are interested in. Existing methods on news recommendation mainly include collaborative filtering methods which rely on direct user-item interactions and content based methods which characterize the content of user reading history. Although these methods have achieved good performances, they still suffer from data sparse problem, since most of them fail to extensively exploit high-order structure information (similar users tend to read similar news articles) in news recommendation systems. In this paper, we propose to build a heterogeneous graph to explicitly model the interactions among users, news and latent topics. The incorporated topic information would help indicate a user’s interest and alleviate the sparsity of user-item interactions. Then we take advantage of graph neural networks to learn user and news representations that encode high-order structure information by propagating embeddings over the graph. The learned user embeddings with complete historic user clicks capture the users’ long-term interests. We also consider a user’s short-term interest using the recent reading history with an attention based LSTM model. Experimental results on real-world datasets show that our proposed model significantly outperforms state-of-the-art methods on news recommendation.     

Hierarchical web resources retrieval by exploiting Fuzzy Formal Concept Analysis

In recent years, knowledge structuring is assuming important roles in several real world applications such as decision support, cooperative problem solving, e-commerce, Semantic Web and, even in planning systems. Ontologies play an important role in supporting automated processes to access information and are at the core of new strategies for the development of knowledge-based systems. Yet, developing an ontology is a time-consuming task which often needs an accurate domain expertise to tackle structural and logical difficulties in the definition of concepts as well as conceivable relationships. This work presents an ontology-based retrieval approach, that supports data organization and visualization and provides a friendly navigation model. It exploits the fuzzy extension of the Formal Concept Analysis theory to elicit conceptualizations from datasets and generate a hierarchy-based representation of extracted knowledge. An intuitive graphical interface provides a multi-facets view of the built ontology. Through a transparent query-based retrieval, final users navigate across concepts, relations and population.     

Jointly modeling intra- and inter-session dependencies with graph neural networks for session-based recommendations

Recently, graph neural networks (GNNs) have achieved promising results in session-based recommendation. Existing methods typically construct a local session graph and a global session graph to explore complex item transition patterns. However, studies have seldom investigated the repeat consumption phenomenon in a local graph. In addition, it is challenging to retrieve relevant adjacent nodes from the whole training set owing to computational complexity and space constraints. In this study, we use a GNN to jointly model intra- and inter-session item dependencies for session-based recommendations. We construct a repeat-aware local session graph to encode the intra-item dependencies and generate the session representation with positional awareness. Then, we use sessions from the current mini-batch instead of the whole training set to construct a global graph, which we refer to as the session-level global graph. Next, we aggregate the K-nearest neighbors to generate the final session representation, which enables easy and efficient neighbor searching. Extensive experiments on three real-world recommendation datasets demonstrate that RN-GNN outperforms state-of-the-art methods.     

A new deep graph attention approach with influence and preference relationship reconstruction for rate prediction recommendation

Graph neural networks have been frequently applied in recommender systems due to their powerful representation abilities for irregular data. However, these methods still suffer from the difficulties such as the inflexible graph structure, sparse and highly imbalanced data, and relatively shallow networks, limiting rate prediction ability for recommendations. This paper presents a novel deep dynamic graph attention framework based on influence and preference relationship reconstruction (DGA-IPR) for recommender systems to learn optimal latent representations of users and items. The entire framework involves a user branch and an item branch. An influence-based dynamic graph attention (IDGA) module, a preference-based dynamic graph attention (PDGA) module, and an adaptive fine feature extraction (AFFE) module are respectively constructed for each branch. Concretely, the first two attention modules concentrate on reconstructing influence and preference relationship graphs, breaking imbalanced and fixed constraints of graph structures. Then a deep feature aggregation block and an adaptive feature fusion operation are built, improving the network depth and capturing potential high-order information expressions. Besides, AFFE is designed to acquire finer latent features for users and items. The DGA-IPR architecture is formed by integrating IDGA, PDGA, and AFFE for users and items, respectively. Experiments reveal the superiority of DGA-IPR over existing recommendation models.     

Informed Patch Enhanced HyperGCN for skeleton-based action recognition

Human skeleton, as a compact representation of action, has attracted numerous research attentions in recent years. However, skeletal data is too sparse to fully characterize fine-grained human motions, especially for hand/finger motions with subtle local movements. Besides, without containing any information of interacted objects, skeleton is hard to identify human–object interaction actions accurately. Hence, many action recognition approaches that purely rely on skeletal data have met a bottleneck in identifying such kind of actions. In this paper, we propose an Informed Patch Enhanced HyperGraph Convolutional Network that jointly employs human pose skeleton and informed visual patches for multi-modal feature learning. Specifically, we extract five informed visual patches around head, left hand, right hand, left foot and right foot joints as the complementary visual graph vertices. These patches often exhibit many action-related semantic information, like facial expressions, hand gestures, and interacted objects with hands or feet, which can compensate the deficiency of skeletal data. This hybrid scheme can boost the performance while keeping the computation and memory load low since only five extra vertices are appended to the original graph. Evaluation on two widely used large-scale datasets for skeleton-based action recognition demonstrates the effectiveness of the proposed method compared to the state-of-the-art methods. Significant accuracy improvements are reported using X-Sub protocol on NTU RGB+D 120 dataset.