Owing to the development of Internet, excessive online data drive users to apply tools to assist them in obtaining desired and useful information. Information retrieval techniques serve as one of the major assistance tools that ease users¡¦ information processing loads. However, most current IR models do not consider processing streaming information which essentially characterizes today¡¦s Web environment. The approach to re-building models based on the full knowledge of data at hand triggered by the new incoming information every time is impractical, inefficient, and costly. Instead, IR models that can be adapted to streaming information incrementally should be considered under the dynamic environment. Therefore, this research is to propose an IR related technique, the incremental aspect model (ISM), which not only uncovers latent aspects from the collected documents but also adapts the aspect model on streaming documents chronologically. There are two stages in ISM: in Stage I, we employ probabilistic latent semantic indexing (PLSI) technique to build a primary aspect model; and in Stage II, with out-of-date data removing and new data folding-in, the aspect model can be expanded using the derived spectral method if new aspects significantly exist. Three experiments are conducted accordingly to verify ISM. Results from the first two experiments show the robust performance of ISM in incremental text clustering tasks. In Experiment III, ISM performs the task of storylines tracking on the 2010 Soccer World Cup event. It illustrates ISM¡¦s incremental learning ability to discover different themes around the event at any time. The feasibility of our proposed approach in real applications is thus justified.

We propose a clustering algorithm which incorporates a similarity-based fuzzy clustering and principal component analysis. The proposed algorithm is capable of discovering clusters with hyper-spherical, hyperellipsoidal, or oblique hyper-ellipsoidal shapes. Besides, the number of the clusters need not be specified in advance by the user. For a given dataset, the orientation, locations, and the number of clusters obtained can truthfully reflect the characteristics of the dataset. Experimental results, obtained by running on datasets generated synthetically, show that our method performs better than other methods.

Linked Data describes an extensive set of structured but heterogeneous datasources where entities are connected by formal semantic descriptions. In thevision of the Semantic Web, these semantic links are extended towards theWorld Wide Web to provide as much machine-readable data as possible forsearch queries. The resulting connections allow an automatic evaluation to findnew insights into the data. Identifying these semantic connections betweentwo data sources with automatic approaches is called link discovery. We derivecommon requirements and a generic link discovery workflow based on similaritiesbetween entity properties and associated properties of ontology concepts. Mostof the existing link discovery approaches disregard the fact that in times ofBig Data, an increasing volume of data sources poses new demands on linkdiscovery. In particular, the problem of complex and time-consuming linkdetermination escalates with an increasing number of intersecting data sources.To overcome the restriction of pairwise linking of entities, holistic clusteringapproaches are needed to link equivalent entities of multiple data sources toconstruct integrated knowledge bases. In this context, the focus on efficiencyand scalability is essential. For example, reusing existing links or backgroundinformation can help to avoid redundant calculations. However, when dealingwith multiple data sources, additional data quality problems must also be dealtwith. This dissertation addresses these comprehensive challenges by designingholistic linking and clustering approaches that enable reuse of existing links.Unlike previous systems, we execute the complete data integration workflowvia a distributed processing system. At first, the LinkLion portal will beintroduced to provide existing links for new applications. These links act asa basis for a physical data integration process to create a unified representationfor equivalent entities from many data sources. We then propose a holisticclustering approach to form consolidated clusters for same real-world entitiesfrom many different sources. At the same time, we exploit the semantic typeof entities to improve the quality of the result. The process identifies errorsin existing links and can find numerous additional links. Additionally, theentity clustering has to react to the high dynamics of the data. In particular,this requires scalable approaches for continuously growing data sources withmany entities as well as additional new sources. Previous entity clusteringapproaches are mostly static, focusing on the one-time linking and clustering ofentities from few sources. Therefore, we propose and evaluate new approaches for incremental entity clustering that supports the continuous addition of newentities and data sources. To cope with the ever-increasing number of LinkedData sources, efficient and scalable methods based on distributed processingsystems are required. Thus we propose distributed holistic approaches to linkmany data sources based on a clustering of entities that represent the samereal-world object. The implementation is realized on Apache Flink. In contrastto previous approaches, we utilize efficiency-enhancing optimizations for bothdistributed static and dynamic clustering. An extensive comparative evaluationof the proposed approaches with various distributed clustering strategies showshigh effectiveness for datasets from multiple domains as well as scalability on amulti-machine Apache Flink cluster.

Entity Resolution (ER) or deduplication aims at identifying entities, such as specific customer or product descriptions, in one or several data sources that refer to the same real-world entity. ER is of key importance for improving data quality and has a crucial role in data integration and querying. The previous generation of ER approaches focus on integrating records from two relational databases or performing deduplication within a single database. Nevertheless, in the era of Big Data the number of available data sources is increasing rapidly. Therefore, large-scale data mining or querying systems need to integrate data obtained from numerous sources. For example, in online digital libraries or E-Shops, publications or products are incorporated from a large number of archives or suppliers across the world or within a specified region or country to provide a unified view for the user. This process requires data consolidation from numerous heterogeneous data sources, which are mostly evolving. By raising the number of sources, data heterogeneity and velocity as well as the variance in data quality is increased. Therefore, multi-source ER, i.e. finding matching entities in an arbitrary number of sources, is a challenging task. Previous efforts for matching and clustering entities between multiple sources (> 2) mostly treated all sources as a single source. This approach excludes utilizing metadata or provenance information for enhancing the integration quality and leads up to poor results due to ignorance of the discrepancy between quality of sources. The conventional ER pipeline consists of blocking, pair-wise matching of entities, and classification. In order to meet the new needs and requirements, holistic clustering approaches that are capable of scaling to many data sources are needed. The holistic clustering-based ER should further overcome the restriction of pairwise linking of entities by making the process capable of grouping entities from multiple sources into clusters. The clustering step aims at removing false links while adding missing true links across sources. Additionally, incremental clustering and repairing approaches need to be developed to cope with the ever-increasing number of sources and new incoming entities. To this end, we developed novel clustering and repairing schemes for multi-source entity resolution. The approaches are capable of grouping entities from multiple clean (duplicate-free) sources, as well as handling data from an arbitrary combination of clean and dirty sources. The multi-source clustering schemes exclusively developed for multi-source ER can obtain superior results compared to general purpose clustering algorithms. Additionally, we developed incremental clustering and repairing methods in order to handle the evolving sources. The proposed incremental approaches are capable of incorporating new sources as well as new entities from existing sources. The more sophisticated approach is able to repair previously determined clusters, and consequently yields improved quality and a reduced dependency on the insert order of the new entities. To ensure scalability, the parallel variation of all approaches are implemented on top of the Apache Flink framework which is a distributed processing engine. The proposed methods have been integrated in a new end-to-end ER tool named FAMER (FAst Multi-source Entity Resolution system). The FAMER framework is comprised of Linking and Clustering components encompassing both batch and incremental ER functionalities. The output of Linking part is recorded as a similarity graph where each vertex represents an entity and each edge maintains the similarity relationship between two entities. Such a similarity graph is the input of the Clustering component. The comprehensive comparative evaluations overall show that the proposed clustering and repairing approaches for both batch and incremental ER achieve high quality while maintaining the scalability.

No description available.