Nowadays, Big Data management has become a key basis for innovation, productivity growth, and competition. The correlated exploitation of data of this magnitude remains primordial to discover valuable insights and support decision making for domains of major interest. Furthermore, despite the complex aspects of Big Data environments, users are usually looking for a unified and appropriate view of this huge and heterogeneous data, to support the extraction of reliable and consistent knowledge. Thus, Big Data integration mechanisms must be considered to provide a uniform query interface, to mediate across large datasets and provide data scientists with a consistent integrated view suitable for analytical exploitations. Thus, this paper presents a semantic-based Big Data integration framework that relies on large-scale ontology matching and probabilistic-logical based assessment strategies. This framework applies optimization mechanisms and leverages parallel-computing paradigms (Hadoop and MapReduce) using commodity computational resources, to efficiently address the Big Data challenges and aspects. Several experiments were conducted and have proven the efficiency of this framework in terms of accuracy, performance, and scalability.

如今，大数据管理已成为创新，生产力增长和竞争的重要基础。对如此大规模的数据进行相关的开发仍然是原始的，以发现有价值的见解并支持重大关注领域的决策。此外，尽管大数据环境复杂，用户通常仍在寻找这种庞大而异构的数据的统一且适当的视图，以支持提取可靠且一致的知识。因此，必须考虑使用大数据集成机制来提供统一的查询界面，以跨大型数据集进行调解，并为数据科学家提供适用于分析开发的一致的集成视图。从而，本文提出了一个基于语义的大数据集成框架，该框架依赖于大规模本体匹配和基于概率逻辑的评估策略。该框架应用优化机制并利用商品计算资源利用并行计算范例（Hadoop和MapReduce），以有效应对大数据挑战和方面。进行了几次实验，并从准确性，性能和可伸缩性方面证明了此框架的效率。