Data & Knowledge Engineering最新文献

This paper presents a plan for preparing for changes in the business environment by analyzing and predicting business district data in Seoul. The COVID-19 pandemic and economic crisis caused by inflation have led to an increase in store closures and a decrease in sales, which has had a significant impact on commercial districts. The number of stores and sales are critical factors that directly affect the business environment and can help prepare for changes. This study conducted correlation analysis to extract factors related to the commercial district’s environment in Seoul and estimated the number of stores and sales based on these factors. Using the Kendaltau correlation coefficient, the study found that existing population and working population were the most influential factors. Linear regression, tensor decomposition, Factorization Machine, and deep neural network models were used to estimate the number of stores and sales, with the deep neural network model showing the best performance in RMSE and evaluation indicators. This study also predicted the number of stores and sales of the service industry in a specific area using the population prediction results of the neural prophet model. The study’s findings can help identify commercial district information and predict the number of stores and sales based on location, industry, and influencing factors, contributing to the revitalization of commercial districts.

With the rapid spread of information, abbreviations are used more and more common because they are convenient. However, the duplication of abbreviations can lead to confusion in many cases, such as information management and information retrieval. The resultant confusion annoys users. Thus, inferring a full name from an abbreviation has practical and significant advantages. The bulk of studies in the literature mainly inferred full names based on rule-based methods, statistical models, the similarity of representation, etc. However, these methods are unable to use various grained contexts properly. In this paper, we propose a flexible framework of Multi-attention mask Abbreviation Context and Full name language model, named MACF to address the problem. With the abbreviation and contexts as the inputs, the MACF can automatically predict a full name by generation, where the contexts can be variously grained. That is, different grained contexts ranging from coarse to fine can be selected to perform such complicated tasks in which contexts include paragraphs, several sentences, or even just a few keywords. A novel multi-attention mask mechanism is also proposed, which allows the model to learn the relationships among abbreviations, contexts, and full names, a process that makes the most of various grained contexts. The three corpora of different languages and fields were analyzed and measured with seven metrics in various aspects to evaluate the proposed framework. According to the experimental results, the MACF yielded more significant and consistent outputs than other baseline methods. Moreover, we discuss the significance and findings, and give the case studies to show the performance in real applications.

Increasingly developing online social networks has enabled users to send or receive information very fast. However, due to the availability of an excessive amount of data in today’s society, managing the information has become very cumbersome, which may lead to the problem of information overload. This highly eminent problem, where the existence of too much relevant information available becomes a hindrance rather than a help, may cause losses, delays, and hardships in making decisions. Thus, in this paper, by defining information overload from a different aspect, we aim to maximize the information propagation while minimizing the information overload (duplication). To do so, we theoretically present the lower and upper bounds for the information overload using a bitwise-based approach as the leverage to mitigate the computation complexities and obtain an approximation ratio of $1-\frac{1}{e}$. We propose two main algorithms, B-square and C-square, and compare them with the existing algorithms. Experiments on two types of datasets, synthetic and real-world networks, verify the effectiveness and efficiency of the proposed approach in addressing the problem.

Keys for graphs are a class of data quality rules that use topological and value constraints to uniquely identify entities in a data graph. They have been studied to support object identification, knowledge fusion, data deduplication, and social network reconciliation. Manual specification and discovery of graph keys is tedious and infeasible over large-scale graphs. To make $\mathrm{GKeys}$ useful in practice, we study the $\mathrm{GKey}$ discovery problem, and present $\mathrm{GKMiner}$, an algorithm that mines keys over graphs. Our algorithm discovers keys in a graph via frequent subgraph expansion, and notably, identifies recursive keys, i.e., where the unique identification of an entity type is dependent upon the identification of another entity type. We introduce the key properties, minimality and support, which effectively help to reduce the space of candidate keys. $\mathrm{GKMiner}$ uses a set of auxillary structures to summarize an input graph, and to identify likely candidate keys for greater pruning efficiency and evaluation of the search space. Our evaluation shows that identifying and using recursive keys in entity linking, lead to improved accuracy, over keys found using existing graph key mining techniques.

Managing unstructured and heterogeneous data, integrating them, and enabling their analysis are among the key challenges in data ecosystems, together with the need to accommodate a progressive growth in these systems by seamlessly supporting extensibility. This is particularly relevant for OLAP analyses on multidimensional cubes stored in data warehouses (DWs), which naturally span large portions of heterogeneous data, possibly relying on different data models (relational, document-based, graph-based). While the management of model heterogeneity in DWs, using for instance multi-model databases, has already been investigated, not much has been done to support extensibility. In a previous paper we have investigated a schema-on-read scenario aimed at granting the extensibility of multidimensional cubes by proposing an architecture to support it and discussing the main open issues associated. This paper takes a step further by presenting xCube, an approach to provide on-demand extensibility of multidimensional cubes in a supply-driven fashion. xCube lets users choose a multidimensional element to be extended, using additional data, possibly uploaded from a data lake. Then, the multidimensional schema is extended by considering the functional dependencies implied by these additional data, and the extended multidimensional schema is made available to users for OLAP analyses. After explaining our approach with reference to a motivating case study in agro-ecology, we propose a proof-of-concept implementation using AgensGraph and Mondrian.

This paper addresses the question of whether and how the development and use of a domain-specific modeling method (DSMM) can increase productivity in the development of technical systems in an industrial setting. This is because an essential prerequisite for DSMMs to become established in operational practice is that productivity increases can be achieved with them and qualitative benefits such as quality assurance, innovation potential, and the like can be exploited. After all, managers’ decisions are ultimately based on whether or not the use of a new method pays off. We illustrate our findings using the example of a DSMM development for the design and realization of electric vehicle testbeds, which we carried out as part of a cooperation project. This work sets the base for possible generalization into other automotive, mechatronic, and technical areas.

Speech Emotion Recognition (SER) is an important area of research in speech processing that aims to identify and classify emotional states conveyed through speech signals. Recent studies have shown considerable performance in SER by exploiting deep contextualized speech representations from self-supervised learning (SSL) models. However, SSL models pre-trained on clean speech data may not perform well on emotional speech data due to the domain shift problem. To address this problem, this paper proposes a novel approach that simultaneously exploits an SSL model and a domain-agnostic spectral feature (SF) through the Mixture of Experts (MoE) technique. The proposed approach achieves the state-of-the-art performance on weighted accuracy compared to other methods in the IEMOCAP dataset. Moreover, this paper demonstrates the existence of the domain shift problem of SSL models in the SER task.

As the development of the Internet and IT technology, short-text based communication is so popular compared with voice based one. Chat-based communication enables rapid, short and massive exchange of message with many people, creates new social problems. ‘Cross-texting’ is one of them. It refers to accidentally sending a text to an unintended person during the concurrent conversations with separated multiple people. Cross-texting would be a serious problem in languages where respectful expressions are required. As text-based communication is getting popular, it is a crucial work to prevent cross-texting by detecting it in advance in languages with honorifics expression such as Korean. In this paper, we proposed two methods detecting a cross-text using a deep learning model. The first model is the formal feature vector, which models dialog by explicitly defining the politeness and completeness features. The second one is the grpah2vec based ChatGram-net model, which models the dialog based on the syllable occurrence relationship. To evaluate the detection performance, we suggest a generating method for cross-text datasets from a actual messenger corpus. In experiment we show that both proposed models detected cross-text effectively, and exceeded the performance of the baseline models.

Relation extraction aims at identifying semantic relations between pairs of named entities from unstructured texts and is considered an essential prerequisite for many downstream tasks in natural language processing (NLP). Owing to the ability in expressing complex relationships and interdependency, graph neural networks (GNNs) have been gradually used to solve the relation extraction problem and have achieved state-of-the-art results. However, the designs of GNN-based relation extraction methods are mostly based on empirical intuition, heuristic, and experimental trial-and-error. A clear understanding of why and how GNNs perform well in relation extraction tasks is lacking. In this study, we investigate three well-known GNN-based relation extraction models, CGCN, AGGCN, and SGCN, and aim to understand the underlying mechanisms of the extractions. In particular, we provide a visual analytic to reveal the dynamics of the models and provide insight into the function of intermediate convolutional layers. We determine that entities, particularly subjects and objects in them, are more important features than other words for relation extraction tasks. With various masking strategies, the significance of entity type to relation extraction is recognized. Then, from the perspective of the model architecture, we find that graph structure modeling and aggregation mechanisms in GCN do not significantly affect the performance improvement of GCN-based relation extraction models. The above findings are of great significance in promoting the development of GNNs. Based on these findings, an engineering oriented MLP-based GNN relation extraction model is proposed to achieve a comparable performance and greater efficiency.

The usefulness of drawing-based psychological testing has been demonstrated in a variety of studies. By using the familiar medium of drawing, drawing-based psychological testing can be applied to a wide range of age groups and is particularly effective with children who have difficulty expressing themselves verbally. Drawing tests are usually implemented face-to-face, requiring specialized counseling staff, and can be time-consuming and expensive to apply to large numbers of children. These problems seem to be solved by applying highly developed artificial intelligence techniques. If artificial intelligence (AI) can analyze children's drawings and perform psychological analysis, it will be possible to use it as a service and take tests online or through smartphones. There have been various attempts to automate the drawing of psychological tests by utilizing deep learning technology to process images. Previous studies using classification have been limited in their ability to extract structural information. In this paper, we analyze the House-Tree-Person Test (HTP), one of the drawing psychological tests widely used in clinical practice, by utilizing object detection technology that can extract more diverse information from images. In addition, we extend the existing research that has been limited to the extraction of relatively simple psychological features and generate a psychological analysis table based on the extracted features that can be used to assist experts in the process of psychological testing. Our research findings indicate that the object detection performance achieves a mean Average Precision (mAP) of approximately 92.6∼94.1 %, and the average accuracy of the psychological analysis table is 94.4 %.