Reasoning over knowledge graphs has attracted considerable attention from researchers and is being widely applied to contribute question answering systems, recommender systems, and other information retrieval systems. However, existing reasoning methods tend to suffer from poor interpretability which is not consistent with human commonsense. The trustworthiness and reliability of the knowledge discover outcomes thus decreased as a result. Inspired by the process of human decision-making, we propose a commonsense-aware hierarchical framework called HDLH, which incorporates commonsense knowledge into hierarchical knowledge graph reasoning process with deep reinforcement learning. HDLH implements hierarchical reasoning process through exploration and exploitation sequentially by applying multi-agent reinforcement learning. Multiple agents in HDLH simulate the multi-level decision-making ability of humans, and reason hierarchically and reasonably to maintain its efficiency and interpretability. Moreover, commonsense knowledge is incorporated by means of the reward-shaping function, ultimately guiding the agent to reason more consistently with human perceptions and reduce the huge search space. We evaluated HDLH with various tasks on five real-world datasets. The experimental results reveal that HDLH achieves better performance compared with state-of-the-art baseline models.

Multi-hop reasoning over knowledge graphs has received plenty of attention from researchers and is being widely applied to facilitate the development of recommender systems, question answering systems, and other information retrieval systems. Existing multi-hop reasoning methods tend to suffer from poor training efficiency as a result of the large search space and have difficulty tackling missing paths during the reasoning process. Accordingly, we propose a sequence-to-sequence model called DT4KGR, in which an encoder–decoder Transformer framework was designed for knowledge graph reasoning. We trained our Transformer model using teacher forcing, so the model processes the entire sequence of reasoning paths in a highly parallel fashion. In this way, faster training speeds are achieved. The model conditions an autoregressive architecture to implement sequence path generation, rather than preceding path exploration, thus performed more robustly to missing paths. We also designed a rule-guided path exploration strategy by combining the local path semantic similarity with the attention mechanism and the global graph information with rule guidance, to sample high-quality training paths for our model. We evaluated DT4KGR through various tasks on seven different real-world datasets. The experimental results reveal DT4KGR achieves better link prediction results compared with the state-of-the-art baseline models while converging 5–8 times faster, especially achieving an approximately 7.8% relative improvement for MRR on NELL-995 dataset. In addition, efficiency studies indicate our model performs better scalability in large-scale knowledge graph environments compared with other methods.

As one of the core technologies of general artificial intelligence, knowledge graph reasoning aims to infer new knowledge from existing knowledge in the knowledge base, providing decision support for knowledge-driven intelligent information services such as information retrieval, question answering, and recommendation systems. However, there are still some issues, such as poor interpretability and low reasoning efficiency, always decrease the current knowledge reasoning performance. To tackle the challenges, this paper proposes a knowledge graph reasoning method LLM4QA, which leverages fine-tuned large language models with chain- of-thought to generate graph query languages SPARQL (i.e., SPARQL Protocol and RDF Query Language) for reasoning. Firstly, an efficient instruction fine-tuning method is applied to fine-tune open-source large language models with chain- of-thought. Then, the fine-tuned open-source large model is used to convert natural language questions into logical forms. Finally, we utilize unsupervised entity relationship retrieval to generate graph database query languages, realizing a natural language knowledge graph question-answering framework. Experimental results demonstrate that this method achieves well performance in terms of inference accuracy and significantly improves model retrieval efficiency.

The objective of document-level relation extraction is to retrieve the relations existing between entities within a document. Currently, deep learning methods have demonstrated superior performance in document-level relation extraction tasks. However, to enhance the model’s performance, various methods directly introduce additional modules into the backbone model, which often increases the number of parameters in the overall model. Consequently, deploying these deep models in resource-limited environments presents a challenge. In this article, we introduce a self-distillation framework for document-level relational extraction. We partition the document-level relation extraction model into two distinct modules, namely, the entity embedding representation module and the entity pair embedding representation module. Subsequently, we apply separate distillation techniques to each module to reduce the model’s size. In order to evaluate the proposed framework’s performance, two benchmark datasets for document-level relation extraction, namely GDA and DocRED are used in this study. The results demonstrate that our model effectively enhances performance and significantly reduces the model’s size.

Entity linking in knowledge-based question answering (KBQA) is intended to construct a mapping relation between a mention in a natural language question and an entity in the knowledge base. Most research in entity linking focuses on long text, but entity linking in open domain KBQA is more concerned with short text. Many recent models have tried to extract the features of raw data by adjusting the neural network structure. However, the models only perform well with several datasets. We therefore concentrate on the data rather than the model itself and created a model DME (Domain information Mining and Explicit expressing) to extract domain information from short text and append it to the data. The entity linking model will be enhanced by training with DME-processed data. Besides, we also developed a novel negative sampling approach to make the model more robust. We conducted experiments using the large Chinese open source benchmark KgCLUE to assess model performance with DME-processed data. The experiments showed that our approach can improve entity linking in the baseline models without the need to change their structure and our approach is demonstrably transferable to other datasets.

User profile plays an important role in analyzing high influencers in the era of big data. For example, the profile can help the political strategists to scrutinize the performance of a candidate or party and overcome their weakness long before election day. In this work, we propose a novel profile construction and application model (PFCA) that aims to discover the overall character of high influencers, and apply PFCA to Donald Trump, one of the most influential people recently, to identify Trump’s comprehensive personalities. We first collect and preprocess profile data from static and dynamic aspects of multi-source data, and then categorize multi-source data topics and emotions. Next, we attempt to construct the profile from four aspects: fundamental characteristic, behavior characteristic, value characteristic and situation characteristic, to explore the correlation among these characteristics. Finally, PFCA is verified by the outcomes of the 2016 and 2020 U.S. presidential elections, and is evaluated by relevant experts, which proves that PFCA is relatively effective and efficient to profile high influencers. Moreover, PFCA can be further applied to related analysis and prediction in the 2024 U.S. presidential election.

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.

Nowadays, 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. However, recently developed reasoning methods usually lack interpretability, and can hardly tackle the large-scale action space problem over KGs. Inspired by the ability of human hierarchical decision making, we propose a multi-hop reasoning framework with deep reinforcement learning (RL) to fill this gap, which incorporates meta information into hierarchical reasoning over KGs. We first use optimization-based meta learning method to initialize parameters for RL agents, allowing for efficient adaptation for tasks in a few gradient steps. Then, a hierarchical RL framework is designed to decompose reasoning tasks into several sub-tasks and solve them separately, performed more efficient and natural than directly solving the entire problem. We further evaluated our model through different tasks on five real world datasets. The experimental results indicate that our method outperforms state-of-the-art baseline models without losing interpretability.

Knowledge Graph (KG) embedding approaches have been proved effective to infer new facts for a KG based on the existing ones–a problem known as KG completion. However, most of them have focused on static KGs, in fact, relational facts in KGs often show temporal dynamics, e.g., the fact (US, has president, Barack Obama, [2009–2017]) is only valid from 2009 to 2017. Therefore, utilizing available time information to develop temporal KG embedding models is an increasingly important problem. In this paper, we propose a new hyperplane-based time-aware KG embedding model for temporal KG completion. By employing the method of time-specific hyperplanes, our model could explicitly incorporate time information in the entity-relation space to predict missing elements in the KG more effectively, especially temporal scopes for facts with missing time information. Moreover, in order to model and infer four important relation patterns including symmetry, antisymmetry, inversion and composition, we map facts happened at the same time into a polar coordinate system. During training procedure, a time-enhanced negative sampling strategy is proposed to get more effective negative samples. Experimental results on datasets extracted from real-world temporal KGs show that our model significantly outperforms existing state-of-the-art approaches for the KG completion task.

In recent years, deep learning models have achieved remarkable progress in the prediction and classification tasks of artificial intelligence systems.However, most of the current deep learning models are black box, which means it is not conducive to human cognitive reasoning process.Meanwhile, with the continuous breakthroughs of artificial intelligence in the researches and applications, high-performance complex algorithms, models and systems generally lack the transparency and interpretability of decision making.This makes it difficult to apply the technologies in a wide range of fields requiring strict interpretability, such as national defense, medical care and cyber security.Therefore, the interpretability of artificial intelligence should be integrated into these algorithms and systems in the process of knowledge reasoning.By means of carrying out explicit explainable intelligence reasoning based on discrete symbolic representation and combining technologies in different fields, a behavior explanation mechanism can be formed which is an important way for artificial intelligence to realize data perception to intelligence perception.A comprehensive review of explainable knowledge graph reasoning was given.The concepts of explainable artificial intelligence and knowledge reasoning were introduced briefly.The latest research progress of explainable knowledge graph reasoning methods based on the three paradigms of artificial intelligence was introduced.Specifically, the ideas and improvement process of the algorithms in different scenarios of explainable knowledge graph reasoning were explained in detail.Moreover, the future research direction and the prospect of explainable knowledge graph reasoning were discussed.