Mining Authoritativeness In Art Historical Photo Archives

In the course of their research, art historians frequently need to refer to historical photo archives when attempting to authenticate works of art. This book, Mining Authoritativeness in Art Historical Photo Archives, provides an aid to retrieving relevant sources and assessing the textual authoritativeness – the internal grounds – of sources of attribution, and to evaluating the authoritativeness of cited scholars. The book aims to do three things: facilitate knowledge discovery in art historical photo archives, support users’ decision-making processes when evaluating contradictory attributions, and provide policies to improve the quality of information in art historical photo archives. The author’s approach is to leverage Semantic Web technologies in order to aggregate, assess, and recommend the most documented authorship attributions. At the same time, the retrieval process allows the providers of art historical data to define a low-cost data integration process with which to update and enrich their collection data. This conceptual framework for assessing questionable information will also be of value to those working in a number of other fields, such as archives, museums, and libraries, as well as to art historians.

Many web pages include structured data in the form of semantic markup, which can be transferred to the Resource Description Framework (RDF) or provide an interface to retrieve RDF data directly. This RDF data enables machines to automatically process and use the data. When applications need data from more than one source the data has to be integrated, and the automation of this can be challenging. Usually, vocabularies are used to concisely describe the data, but because of the decentralized nature of the web, multiple data sources can provide similar information with different vocabularies, making integration more difficult. This book, Multi-modal Data Fusion based on Embeddings, describes how similar statements about entities can be identified across sources, independent of the vocabulary and data modeling choices. Previous approaches have relied on clean and extensively modeled ontologies for the alignment of statements, but the often noisy data in a web context does not necessarily adhere to these prerequisites. In this book, the use of RDF label information of entities is proposed to tackle this problem. In combination with embeddings, the use of label information allows for a better integration of noisy data, something that has been empirically confirmed by experiment. The book presents two main scientific contributions: the vocabulary and modeling agnostic fusion approach on the purely textual label information, and the combination of three different modalities into one multi-modal embedding space for a more human-like notion of similarity. The book will be of interest to all those faced with the problem of processing data from multiple web-based sources.

The distributed setting of RDF stores in the cloud poses many challenges, including how to optimize data placement on the compute nodes to improve query performance. In this book, a novel benchmarking methodology is developed for data placement strategies; one that overcomes these limitations by using a data-placement-strategy-independent distributed RDF store to analyze the effect of the data placement strategies on query performance. Frequently used data placement strategies have been evaluated, and this evaluation challenges the commonly held belief that data placement strategies which emphasize local computation lead to faster query executions. Indeed, results indicate that queries with a high workload can be executed faster on hash-based data placement strategies than on, for example, minimal edge-cut covers. The analysis of additional measurements indicates that vertical parallelization (i.e., a well-distributed workload) may be more important than horizontal containment (i.e., minimal data transport) for efficient query processing. Two such data placement strategies are proposed: the first, found in the literature, is entitled overpartitioned minimal edge-cut cover, and the second is the newly developed molecule hash cover. Evaluation revealed a balanced query workload and a high horizontal containment, which lead to a high vertical parallelization. As a result, these strategies demonstrated better query performance than other frequently used data placement strategies. The book also tests the hypothesis that collocating small connected triple sets on the same compute node while balancing the amount of triples stored on the different compute nodes leads to a high vertical parallelization.

Social robots are embodied agents that perform knowledge-intensive tasks involving several kinds of information from different heterogeneous sources. This book, Engineering Background Knowledge for Social Robots, introduces a component-based architecture for supporting the knowledge-intensive tasks performed by social robots. The design was based on the requirements of a real socially-assistive robotic application, and all the components contribute to and benefit from the knowledge base which is its cornerstone. The knowledge base is structured by a set of interconnected and modularized ontologies which model the information, and is initially populated with linguistic, ontological and factual knowledge retrieved from Linked Open Data. Access to the knowledge base is guaranteed by Lizard, a tool providing software components, with an API for accessing facts stored in the knowledge base in a programmatic and object-oriented way. The author introduces two methods for engineering the knowledge needed by robots, a novel method for automatically integrating knowledge from heterogeneous sources with a frame-driven approach, and a novel empirical method for assessing foundational distinctions over Linked Open Data entities from a common-sense perspective. These effectively enable the evolution of the robot’s knowledge by automatically integrating information derived from heterogeneous sources and the generation of common-sense knowledge using Linked Open Data as an empirical basis. The feasibility and benefits of the architecture have been assessed through a prototype deployed in a real socially-assistive scenario, and the book presents two applications and the results of a qualitative and quantitative evaluation.

The vast amount of data available on the web has led to the need for effective retrieval techniques to transform that data into usable machine knowledge. But the creation of integrated knowledge, especially knowledge about the same entity from different web data sources, is a challenging task requiring the solving of interoperability problems. This book addresses the problem of knowledge retrieval and integration from heterogeneous web sources, and proposes a holistic semantic knowledge retrieval and integration approach to creating knowledge graphs on-demand from diverse web sources. Semantic Web Technologies have evolved as a novel approach to tackle the problem of knowledge integration from heterogeneous data, but because of the Extraction-Transformation-Load approach that dominates the process, knowledge retrieval and integration from web data sources is either expensive, or full physical integration of the data is impeded by restricted access. Focusing on the representation of data from web sources as pieces of knowledge belonging to the same entity which can then be synthesized as a knowledge graph helps to solve interoperability conflicts and allow for a more cost-effective integration approach, providing a method that enables the creation of valuable insights from heterogeneous web data. Empirical evaluations to assess the effectiveness of this holistic approach provide evidence that the methodology and techniques proposed in this book help to effectively integrate the disparate knowledge spread over heterogeneous web data sources, and the book also demonstrates how three domain applications of law enforcement, job market analysis, and manufacturing, have been developed and managed using the approach.

Linked Data is a method of publishing structured data to facilitate sharing, linking, searching and re-use. Many such datasets have already been published, but although their number and size continues to increase, the main objectives of linking and integration have not yet been fully realized, and even seemingly simple tasks, like finding all the available information for an entity, are still challenging. This book, Services for Connecting and Integrating Big Numbers of Linked Datasets, is the 50th volume in the series ‘Studies on the Semantic Web’. The book analyzes the research work done in the area of linked data integration, and focuses on methods that can be used at large scale. It then proposes indexes and algorithms for tackling some of the challenges, such as, methods for performing cross-dataset identity reasoning, finding all the available information for an entity, methods for ordering content-based dataset discovery, and others. The author demonstrates how content-based dataset discovery can be reduced to solving optimization problems, and techniques are proposed for solving these efficiently while taking the contents of the datasets into consideration. To order them in real time, the proposed indexes and algorithms have been implemented in a suite of services called LODsyndesis, in turn enabling the implementation of other high level services, such as techniques for knowledge graph embeddings, and services for data enrichment which can be exploited for machine-learning tasks, and which also improve the prediction of machine-learning problems.

Most people need textual or visual interfaces to help them make sense of Semantic Web data. In this book, the author investigates the problems associated with generating natural language summaries for structured data encoded as triples using deep neural networks. An end-to-end trainable architecture is proposed, which encodes the information from a set of knowledge graph triples into a vector of fixed dimensionality, and generates a textual summary by conditioning the output on this encoded vector. Different methodologies for building the required data-to-text corpora are explored to train and evaluate the performance of the approach. Attention is first focused on generating biographies, and the author demonstrates that the technique is capable of scaling to domains with larger and more challenging vocabularies. The applicability of the technique for the generation of open-domain Wikipedia summaries in Arabic and Esperanto – two under-resourced languages – is then discussed, and a set of community studies, devised to measure the usability of the automatically generated content by Wikipedia readers and editors, is described. Finally, the book explains an extension of the original model with a pointer mechanism that enables it to learn to verbalise in a different number of ways the content from the triples while retaining the capacity to generate words from a fixed target vocabulary. The evaluation of performance using a dataset encompassing all of English Wikipedia is described, with results from both automatic and human evaluation both of which highlight the superiority of the latter approach as compared to the original architecture.