Citation Context

...tching that first builds a dependency graph for the attributes in each data source by mining the instance values, and then uses a graph matching algorithm to find attribute correspondences. Dai et al =-=[14]-=- propose an information-theoretic measure to validate the matching results that can work even if both schema information and instance values do not match (e.g., merging of two customer databases from ...

Citation Context

...a. In general, viewing information theory as a tool to express and quantify notions of information content and information transfer has been very successful as a way of extracting structure from data =-=[14, 3, 9, 5, 7, 8, 2]-=-. In this tutorial, we will explore the use of information theory as part of a data representation and analysis toolkit. We will do this with illustrative examples that span a wide range of topics of ...

Citation Context

...8.45 0.42 9.42 9.61 6.03 200 0.03 3158.37 4.00 27.03 24.59 17.17 300 0.03 - 8.74 46.88 47.72 31.02 400 0.05 - 24.91 73.50 65.81 32.83 500 0.08 - 41.78 321.56 100.39 58.55 a schema matching context is =-=[12]-=-, where the authors construct signatures for validating rather than inferring schema matches. That work concentrates on information theoretic principles only (using IB). In the present work we conside...

Citation Context

...ed with the K clusters. In the second step, we measure the distance between tags by comparing the probability distribution of their associated membership to each of the K clusters using KL-divergence =-=[8, 9]-=-. KL-divergence is a function for measuring the relative entropy between two distributions. Specifically, if the probabilistic cluster membership distributions for two tags are p and q respectively, t...

Citation Context

... obtain a similarity value (from now on the words column and attribute are used interchangeably). An N-gram is simply a substring of length N consisting of contiguous characters. In particular, using distinct types we strive to capture patterns from both the raw text of GIS datasets and encoded versions of this text which substitute the individual letters for their character types (i.e., letters are replaced by an ‘a’, numbers are replaced by a ‘n’, etc.) A number of schema matching publications describing methods tailored more to the database community influenced our work. Dai, Koudas et al. [5] discussed content-based schema matching based on distributions of N-grams among compared columns. Despite the influence of this publication, some crucial differences exist between their approach and the methods explained here. First, their approach used data sources containing raw text from “Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to rep...

Citation Context

...l. [11], so we summarize those additional metrics to complement the quality measures introduced above: • response time, refers to the running time required by the matchers to complete a matching task =-=[2, 6, 7, 5, 35, 21, 10]-=-. • matching accuracy, is defined as the percentage of source/target attributes that are matched correctly [13, 8]. While designing MatchBench, we decided to use some of the evaluation criteria introd...

Citation Context

...termediate schema for thissexample is: Gene (GID, Name, Organism, CHRO, Ref, Nucl, Strend, Symbol, Comments).sValidation of schema matching is an essential task in the process of database integration =-=[41]-=-. Insthe following, we show that the properties of the input schemas as well as the characteristics of thesrelational model are maintained when input schemas are integrated into a global schema in our...

Citation Context

...IS ontologies based on similarity calculations involving data types from the analysis of specific instances [38]. Dai et al. describe the importance of an “invalidation certificate” as a decision aid =-=[21]-=-. The semanticmatch between attributes of different database was tested and when appropriate, an invalidation certificate was issued to the operator [21]. Validity is difficult to demonstrate reliably...