Abstract With the growing availability of different knowledge graphs in a variety domains, question answering over graph (KG-QA) becomes prevalent information retrieval approach. Current KG-QA methods usually resort to semantic parsing, search or neural matching models. However, they cannot well tackle increasingly long input questions and complex needs. In this work, we propose new approach, leveraging rich domain context graph. We incorporate approach with answer descriptions....

This paper concerns with the actual problems in legal work. We apply event extraction technology to case description part Chinese text. define type, argument and role of larceny case, construct a dataset through data annotation. divide into two steps: trigger word joint assignment. use BERT obtain character vectors, BiLSTM-CRF model for at first step, combine additional features results then input them CRF second step an improvement result. display extracted information time series realize...

In this paper, we investigate the classification of microscopic tumours using full digital mammography images. Firstly, to address shortcomings traditional image segmentation methods, two different deep learning methods are designed achieve uterine fibroids. The lab model is used optimize lesion edge detailed information by void convolution algorithm and fully connected CRF, semantic networks compared obtain best results. Mask RCNN case effectively extract features through ResNet structure,...

The "arms race" of Large Language Models (LLMs) demands novel, challenging, and diverse benchmarks to faithfully examine their progresses. We introduce GraphArena, a benchmarking tool designed evaluate LLMs on graph computational problems using million-scale real-world graphs from scenarios such as knowledge graphs, social networks, molecular structures. GraphArena offers suite 10 tasks, encompassing four polynomial-time (e.g., Shortest Distance) six NP-complete challenges Travelling...

In the context of digital transformation and economic globalization, R&D (research development) internationalization is essential for enterprises to utilize global resources achieve technological innovation. This study examines Chinese A-share-listed industrial companies with active overseas from 2010 2022 using a Poisson panel fixed-effects model assess how influences internationalization. The findings confirm that significantly enhances depth breadth internationalization, even when...

Large Language Models (LLMs) have demonstrated strong reasoning abilities, making them suitable for complex tasks such as graph computation. Traditional steps paradigm problems is hindered by unverifiable steps, limited long-term reasoning, and poor generalization to variations. To overcome these limitations, we introduce GCoder, a code-based LLM designed enhance problem-solving in generalized computation problems. Our method involves constructing an extensive training dataset, GraphWild,...

Let p be a prime number. In this paper, the author sets up canonical correspondence between polynomial functions over ℤ/p 2 ℤ and 3-tuples of ℤ/pℤ. Based on correspondence, he proves reproves some fundamental results permutation polynomials mod l . The main new result is characterization strong orthogonal systems ℤ.

We explore the use of policy approximations to reduce computational cost learning Nash equilibria in zero-sum stochastic games. propose a new Q-learning type algorithm that uses sequence entropy-regularized soft policies approximate during Q-function updates. prove under certain conditions, by updating regularized Q-function, converges equilibrium. also demonstrate proposed algorithm's ability transfer previous training experiences, enabling agents adapt quickly environments. provide dynamic...

The task of named entity recognition is to identify certain types entities with special meanings from the text. It a basic in natural language processing and foundation higher-level tasks such as relation extraction, knowledge graph, question answering system. correctness has huge influence on effectiveness upper layer application. This paper mainly studies problem Chinese medical field. By extracting information about disease insurance text labeling disease, treatment, symptom, data set for...

Graph embedding has been widely used to process various downstream tasks on large-scale graphs, i.e. node classification, community detection and link prediction. Among methods, Deep Infomax (DGI) is a newly proposed method which achieves excellent performance in classification. However, such outstanding achievement may cause the over-mining issue of user privacy robustness this still unexplored. In paper, we investigate how disturb classification accuracy DGI from an attacker's perspective....

We explore the use of policy approximations to reduce computational cost learning Nash equilibria in zero-sum stochastic games. propose a new Q-learning type algorithm that uses sequence entropy-regularized soft policies approximate during Q-function updates. prove under certain conditions, by updating entropy regularization, converges equilibrium. also demonstrate proposed algorithm's ability transfer previous training experiences, enabling agents adapt quickly environments. provide dynamic...

Abstract Using the ergodicity of chaotic iterative sequences to realize data compression is a new research perspective. We find that, under suitable initial conditions, one or more local segments that are numerically identical random integer appear on sequence, which prerequisite for compression. Based this, this paper designs lossless method based three-dimensional product-triangular sequences. The proposed only needs input small amount information compress large through an sequence limited...

In contemporary society, men gradually increase their demand for shavers to keep faces clean. Women also seem pursue long hair by using a wig or artificial integrations. So, this paper will analyze these phenomena studying the role of male beard and female in mating selection from an evolutionary psychology perspective. This work is largely based on empirical evidence, previous studies , relevant theories (such as short- long-term strategies, parental investment theory, good genes theory)...

Language models have been increasingly popular for automatic creativity assessment, generating semantic distances to objectively measure the quality of creative ideas. However, there is currently a lack an assessment system evaluating ideas in Chinese language. To address this gap, we developed TransDis, scoring using transformer-based language models, capable providing valid originality (quality) and flexibility (variety) scores Alternative Uses Task (AUT) responses Chinese. Study 1...

Abstract Topological indices are useful molecular descriptors to measure Quantitative Structure-Activity Relationship (QSAR), Structure-Property (QSPR) and Structure-Toxicity (QSTR). Recently, some novel topological of graphs, called the first Gourava index ($CO_1$) second ($CO_2$), have been proposed characterize nature chemical compounds or interconnection networks. The graph $G$ denoted by $C{O_1}(G) =\sum\limits_{uv \in E(G)} {[{d_u}} + {d_v} {d_u}{d_v}]$ $C{O_2}(G) = \sum\limits_{uv...

In the blast furnace iron-making, condition is directly related to production efficiency and life. Currently, analysis of mainly based on artificial experience. Further progress does not exist in due difficulty mechanism modeling. this paper, biclustering algorithm applied condition, which more suitable for high dimensional data has ability deal with large-scale data. First, rank Order-Preserving Submatrix (r-OPSM) hierarchical search proposed, improves robustness OPSM noise used correlation...