Knowledge graphs and ontologies play an essential role in integrating, standardizing, and reasoning about complex data across domains. In recent studies, leveraging knowledge graphs in AI use cases, instead of traditional relational databases, led to quality improvements by up to 38 percentage points. However, learning ontologies from relational databases remains a challenging task due to the impedance mismatch between both modeling concepts. An understanding of which ontology learning system performs best, and why, is missing, as no established benchmark exists. We present BURR, a benchmark for evaluating ontology learning systems from relational databases. To evaluate the ontology learning space, we introduce a novel mapping-based metric and provide a comprehensive benchmark data collection. This collection of 54 scenarios consists of real-world database-ontology mappings, including industry data, and of a micro-benchmark evaluating the behavior of systems in encapsulated scenarios. We demonstrate the applicability of BURR by evaluating widely used ontology learning systems, including traditional rule-based as well as LLM-based approaches, on the benchmark. The results emphasize the current strengths of simple rule-based approaches compared to LLM-based systems, while also highlighting the significant research potential of LLMs in ontology learning.

The amount of publicly available Spatio-temporal (ST) data is growing daily and possesses an increasing degree of complexity in more and more use cases. Besides spatial queries such as intersection, the requirements of current applications like Digital Twins (DT) go beyond the limits of a single data processing platform and need to combine a variety of queries with filtering ( e.g., k -NN), aggregation (e.g., counting), ranking (e.g., page-rank), clustering (e.g., k-means, ST-DBSCAN) and more, on ST-models. Since existing ST-platforms are highly specialized for a subset of these operations, it seems logical to distribute the data and queries across several of these systems. However, efficient p rocessing a cross d ifferent s ystems i s still a major challenge in polyglot data management and often demands manual query planning. To solve the automatic planning of those complex queries, we present an approach for cross-platform processing of ST-tasks that uses a symmetric join to handle platform heterogeneity and includes a novel algorithm for operator placement based on a latency model. Although the underlying problem is NP-hard and additional network transfers slow down the overall processing time, experiments on real-world tasks for DTs have shown that cross-platform processing can speed up well-known ST-tasks compared to the expensive query reformulations performed by state-of-the-art ST single-platform solutions.

Spatio-temporal trajectories are essential factors for systems used in public transport, social ecology, and many other disciplines where movement is a relevant dynamic process. Each trajectory describes multiple state changes over time, induced by individual decision-making, based on psychological and social factors with physical constraints. Since a crucial factor of such systems is to reason about the potential trajectories in a closed environment, the primary problem is the realistic replication of individual decision making. Mental factors are often uncertain, not available or cannot be observed in reality. Thus, models for data generation must be derived from abstract studies using probabilities. To solve these problems, we present Multi-Agent-Trajectory-Learning (MATL), a state transition model to learn and generate human-like Spatio-temporal trajectory data. MATL combines Generative Adversarial Imitation Learning (GAIL) with a simulation system that uses constraints given by an agent-based model (Aℬℳ). We use GAIL to learn policies in conjunction with the Aℬℳ, resulting in a novel concept of individual decision making. Experiments with standard trajectory predictions show that our approach produces similar results to real-world observations.

"Bad" data has a direct impact on 88% of companies, with the average company losing 12% of its revenue due to it. Duplicates - multiple but different representations of the same real-world entities - are among the main reasons for poor data quality, so finding and configuring the right deduplication solution is essential. Existing data matching benchmarks focus on the quality of matching results and neglect other important factors, such as business requirements. Additionally, they often do not support the exploration of data matching results. To address this gap between the mere counting of record pairs vs. a comprehensive means to evaluate data matching solutions, we present the Frost platform. It combines existing benchmarks, established quality metrics, cost and effort metrics, and exploration techniques, making it the first platform to allow systematic exploration to understand matching results. Frost is implemented and published in the open-source application Snowman, which includes the visual exploration of matching results, as shown in Figure 1.

Due to the increasing variety of the current database landscape, polyglot data management has become a hot research topic in recent years. The underlying idea is to combine the benefits of different data stores behind a predefined set of common interfaces and thus address use cases that individual stores cannot meet. This can be accomplished using different approaches which vary greatly in terms of capabilities, functionality, and architectural concepts. This tutorial provides a detailed overview of the current state of research in polyglot data management. We motivate its use by showing the high diversity of existing data stores and discussing three use cases in which individual stores are insufficient. Thereafter, we present different taxonomies for classifying polyglot data systems and give a detailed review of a number of selected systems. Finally, we compare these systems based on their features and discuss open challenges that still need to be addressed in future research.

Content delivery networks (CDNs) are critical for minimizing access latency in the Web as they efficiently distribute online resources across the globe. But since CDNs can only be enabled on the scope of entire websites (and not for individual users or user groups), the effects of page speed acceleration are often quantified with potentially skewed before-after comparisons rather than statistically sound A/B tests. We introduce the system Beaconnect for collecting and analyzing Web performance data without being subject to these limitations. Our contributions are threefold. First, Beaconnect is natively compatible with A/B testing Web performance as it is built for a custom browser-based acceleration approach and thus does not rely on traditional CDN technology. Second, we present our continuous aggregation pipeline that achieves sub-minute end-to-end latency. Third, we describe and evaluate a scheme for continuous real-time reporting that is especially efficient for large customers and processes data from over 100 million monthly users at Baqend.

The detection of duplicates is an essential task in data cleaning and integration and has steadily gained importance especially for researchers and practitioners that need to process and integrate large volumes of potentially unclean data on a daily basis. To evaluate the quality and performance of duplicate detection algorithms, labeled test data are required that provide information on the contained duplicates. Current approaches for generating test data, however, are either not scalable (and therefore limited to small datasets) or not able to generate realistic data values and errors, especially outdated values. In this paper, we propose a scheme for generating test datasets that addresses both these issues and present a test dataset generated with it. Our approach relies on using historical data from the North Carolina voter register which (1) is realistic as it contains actual voter data and (2) facilitates generating realistic duplicates through the fact that current data values were collected at every election through manually filled out applications. The generated test dataset comprises more than 120 million records with up to 90 attribute values each. To the best of our knowledge, we are the first who providerealistic test data for duplicate detection at this scale.

Duplicate detection identifies multiple records in a dataset that represent the same real-world object. Many such approaches exist, both in research and in industry. To investigate essential properties of duplicate detection algorithms, such as their result quality or runtime behavior, they must be executed on suitable test data. The quality evaluation requires that these test data are labeled, constituting a ground truth. Correctly labeled, sizable, and real or at least realistic test datasets, however, are not easy to obtain, creating an obstacle for the advancement of research. In this tutorial, we present common methods to evaluate duplicate detection algorithms and to generate labeled test data. We close with a discussion of open problems.