No abstract

Entity Resolution (ER) is an important real world problem that has attracted significant research interest over the past few years. It deals with determining which object descriptions co-refer in a dataset. Due to its practical significance for data mining and data analysis tasks many different ER approaches has been developed to address the ER challenge. This paper proposes a new ER Ensemble framework. The task of ER Ensemble is to combine the results of multiple base-level ER systems into a single solution with the goal of increasing the quality of ER. The framework proposed in this paper leverages the observation that often no single ER method always performs the best, consistently outperforming other ER techniques in terms of quality. Instead, different ER solutions perform better in different contexts. The framework employs two novel combining approaches, which are based on supervised learning. The two approaches learn a mapping of the clustering decisions of the base-level ER systems, together with the local context, into a combined clustering decision. The paper empirically studies the framework by applying it to different domains. The experiments demonstrate that the proposed framework achieves significantly higher disambiguation quality compared to the current state of the art solutions.

Entity resolution is a very common Information Quality (IQ) problem with many different applications. In digital libraries, it is related to problems of citation matching and author name disambiguation; in Natural Language Processing, it is related to coreference matching and object identity; in Web application, it is related to Web page disambiguation. The problem of Entity Resolution arises because objects/entities in real world datasets are often referred to by descriptions, which might not be unique identifiers of these entities, leading to ambiguity. The goal is to group all the entity descriptions that refer to the same real world entities. In this paper we present a graphical approach for entity resolution. It complements the traditional methodology with the analysis of the entity-relationship graph constructed for the dataset being analyzed. The paper demonstrates that a technique that measures the degree of interconnectedness between various pairs of nodes in the graph can significantly improve the quality of entity resolution. Furthermore, the paper presents an algorithm for making that technique self-adaptive to the underlying data, thus minimizing the required participation from the domain-analyst and potentially further improving the disambiguation quality.

Nowadays, searches for the web pages of a person with a given name constitute a notable fraction of queries to Web search engines. Such a query would normally return web pages related to several namesakes, who happened to have the queried name, leaving the burden of disambiguating and collecting pages relevant to a particular person (from among the namesakes) on the user. In this paper, we develop a Web People Search approach that clusters web pages based on their association to different people. Our method exploits a variety of semantic information extracted from web pages, such as named entities and hyperlinks, to disambiguate among namesakes referred to on the web pages. We demonstrate the effectiveness of our approach by testing the efficacy of the disambiguation algorithms and its impact on person search.

Data mining practitioners frequently have to spend significant portion of their project time on data preprocessing before they can apply their algorithms on real-world datasets. Such a preprocessing is required because many real-world datasets are not perfect, but rather they contain missing, erroneous, duplicate data and other data cleaning problems. It is a well established fact that, in general, if such problems with data are not corrected, applying data mining algorithm can lead to wrong results. The latter is known as the "garbage in, garbage out" principle. Given the significance of the problem, numerous data cleaning techniques have been designed in the past to address the aforementioned problems with data.In this paper, we address one of the data cleaning challenges, called object consolidation. This important challenge arises because objects in datasets are frequently represented via descriptions (a set of instantiated attributes), which alone might not always uniquely identify the object. The goal of object consolidation is to correctly consolidate (i.e., to group/determine) all the representations of the same object, for each object in the dataset. In contrast to traditional domain-independent data cleaning techniques, our approach analyzes not only object features, but also additional semantic information: inter-objects relationships, for the purpose of object consolidation. The approach views datasets as attributed relational graphs (ARGs) of object representations (nodes), connected via relationships (edges). The approach then applies graph partitioning techniques to accurately cluster object representations. Our empirical study over real datasets shows that analyzing relationships significantly improves the quality of the result.