Philip Schreur, Stanford University, USA
Modeling Events as Linked Data: BIBFRAME, Performed Music, and Audiovisual Materials [presentation ppt]
Abstract: In 2012, the Library of Congress (LC) began a project to end libraries isolation from the semantic web through the creation of a new, linked-data based communication format, called BIBFRAME, as a replacement for the MARC formats. The development of BIBFRAME has been a complex one as its creators try to balance the need to capture the data encoded in MARC, the constraints of RDF, and input from the community it hopes to serve. The community is a multifaceted one, however. One of the most difficult to serve has been the music/audiovisual community whose resources are rooted in events. A recent modeling study by the Audiovisual Section of LC extends the BIBFRAME model to better incorporate the needs of those communities. The focus on work and event as cofactors in data creation provides a far superior model for resource discovery.
1. Jian Qin, Syracuse University, USA
Turning Data into Knowledge: Modeling and Structuring for Linked Open Data [presentation ppt]
Abstract: This presentation will discuss the new challenges in knowledge organization through an example of turning the data in a relational database into a graph database. The example involves a large collection of metadata from the GenBank, an international data repository for curating genetic sequences. In the metadata records contained are author names, information about the genetic sequence, taxonomic classes to which the sequence belongs, publications related to the sequence, and authors and institutions responsible for the submissions. All the metadata records are stored in a relational database. How to turn data into knowledge is one of the questions we faced. Converting the data in a relational database into a graph database is the strategy we used for accomplishing the goal. This process requires knowledge modeling and organization and offers some useful insights into knowledge organization in the data-driven environment. Although modeling data properties and relationships between properties is not new in computer and information science, it is new in the sense of the object being modeled. In this case, it is the data that is being modeled and converted into a knowledge structure.
2. Sam Oh, Seonghun Kim, Jinhyeong Lee, SKKU iSchool, Seoul, Korea
A role of Ontology in Social data Analytics [presentation ppt]
Abstract: Ontology-supported retrieval systems have been constructed on the basis of Topic Maps and RDF/OWL technology to improve how users query, navigate, access, and interact with data and knowledge in specific disciplines. This study builds on this research thread by constructing RDF- and Topic Maps-based interfaces for querying and browsing Wikipedia. The effectiveness of retrieval using the two systems, as compared to the existing Wikipedia service, is assessed through user evaluations according to uniform objective and subjective criteria. Wikipedia was chosen as the focus for this study because it is currently the Web encyclopedia service of choice; because relationships between items are recorded in minute detail; and because all content can be re-used free of charge. Semantic query-based systems can be implemented for Wikipedia content relatively easily on the basis of RDF datasets provided by the DBpedia project. An objective evaluation of effectiveness measured quantifiable components such as retrieval time, retrieval iterations, and page views, showing statistically significant differences in retrieval time among the three systems for all query types. A subjective evaluation used questionnaires to assess the accuracy, usability, usefulness, and degree of satisfaction of users for each system, showing significant differences in all query types, including system accuracy, usability, usefulness, and degree of user satisfaction.
3. Marcia, Zeng, Julaine Clunis, Kent State University, USA; Vlasimir Alexiev, Ontotext, Bulgaria
Innovative Use of Knowledge Organization Systems (KOS) that are Published as Linked Open Data (LOD) [presentation ppt]
Abstract: The KOS creations involve tremendous intellectual efforts and human resources. In the Linked Open Data (LOD) environment, many KOS resources have become publicly available while easy to obtain and reuse. This presentation intends to share the findings of a research conducted by the authors and bring both developers and potential users together to think how to build on the existing LOD KOS and maximize their usages in innovative ways. The innovative uses are summarized according to the functionalities of LOD KOS, including:
- I. For vocab creators/managers: (a) as the resources of creating, maintaining, enriching, extending, and translating a controlled vocabulary; and (b) as the vocabulary management facility;
- ||. For LOD data creators: (a) as the source of http URIs/IRIs used in data-transformation that enable owners of structured data to convert and publish their metadata under the Linked Data principles (i.e., use HTTP URIs/IRIs as names of things); (b) as the source of http URIs/IRIs in creating machine-understandable data for their websites by individuals and web masters; and (c) as components and plug-ins in editing and publishing workflows or for website products;
- III. For Web site developers, (a) as web services for resource discovery in the Web of Things (e.g., used for semantic enrichment procedure and for multilingual and trans-lingual searching); (b) as middleware for visualizing concepts and relationships, mashing-up resources, and providing user-friendly knowledge presentation; and
- IV. For Researchers, as knowledge bases of research, for example, for obtaining special graphs or datasets for very complicated questions, and revealing unknown relationships.
4. Wnhong Jang and Sam Oh, SKKU iSchool, Seoul, Korea
Leveraging semantic structure to improve retrieval: a comparative evaluation of result-set browsing based on topic maps and RDF ontologies [presentation ppt]
Abstract: As social data becomes richer and diverse, the research in the field of sentiment analysis (also known as opinion mining) of social data has grown rapidly. This talk will present our recent project with LG CNS (a leading advanced analytics company in Korea) that seeks to improve sentiment analysis by modeling vocabulary with SKOS and RDF/OWL. SKOS provides interoperability and reuse open data vocabulary. This project has the following goals to secure interoperability and utilize open data vocabulary:
- Building a SKOS/OWL-based vocabulary/ontology management system on the domain of smartphones
- Expanding current data visualization on the smartphone domain using D3.js
- Interviewing ontology designers and social data analysts to see their reactions to the pilot system.
5. Ying-Hsang Liu and Philip Hider, Charles Sturt University, Australia
Retrieval Effectiveness of Uncontrolled and Controlled Indexing Vocabulary in Education Database
Abstract: The project was commissioned by the Office for Learning & Teaching (OLT) in Australia to investigate how the various existing and prospective resources in the Resource Library (http://www.olt.gov.au/resource-library) should be described and indexed, so that they can be accessed more effectively. The project was to address concerns in retrieval effectiveness by implementing its findings through a reindexing of the Resource Library database and providing guidelines for ongoing indexing, as new resources were entered into the content management system.
Two professional indexers in the education domain participated our study as searchers and judges of document relevance. Two information retrieval experiments were conducted before and after the reindexing of the database. A total of 1,430 documents from the de-duplicated list were assessed by the two judges. The inter-rater agreement, Cohen’s kappa coefficient was 0.91, indicating a very good degree of agreement between the two judges.
Our results reveal that the average precision for initial searches across all topics was 0.33 (n = 80), whereas the average recall was 0.45 (n = 68) before the reindexing. The result was an increase in average recall from 25.1% to 37.1% and an increase in average precision from 37.6% to 50.4% after the reindexing.
Our findings provide empirical evidence of the effectiveness of controlled indexing vocabulary in the education domain. The research findings have implications for the question of whether it's worthwhile expending the resources to create and maintain controlled indexing vocabulary for NKOS (Networked Knowledge Organization Systems) development and digital libraries.
6. Robert Allen, Yonsei University, South Korea
Repositories with Direct Representation [paper pdf]
Abstract: We propose that a new generation of digital repositories should be based on direct representation of the contents with rich semantics and models
rather than as collections of documents. The contents of such repositories would be highly structured. These structures should allow the users of the repositories to focus on meaningful relationships of the contents of the repositories. These repositories would apply and extend our earlier proposals for model-oriented information organization. They could build on the semantic frameworks which are currently being developed. For instance, direct repositories of scholarly communication in biology might be based on current work on upper ontologies if those are extended to more explicitly cover state changes, instances, and scenarios. Other approaches to rich semantics and modeling which are based on insights from object-oriented design and frame semantics also need to be explored. In addition to basic semantics, the representations need to allow for discourse and other repository knowledge support services. For instance, they would need to continually and systematically update the semantic framework as research findings and new theories reshape the structure of the knowledge base.
7. Maja Žumer, University of Ljubljana, Slovenia
Modelling versions of KOS with FRBR [presentation ppt]
Abstract: Functional Requirements for Bibliographic Records (FRBR) family of models allows us to capture versions of KOS which are result of updating, abridgement or translation. Describing KOS at different levels of abstractions (work, expression, manifestation, item; often shortened as WEMI) enables meaningful clustering of versions and recording explicit relationships among them, for example, which version was used as the source of translation. It is therefore possible to explicitly track the derivation chain of multilingual/multicultural KOS. The new harmonized model, now tentatively called FRBR LRM (Library Reference Model) is now a draft and is being discussed within the IFLA community. It will be presented in the context of modelling both versions of KOS and the subject relationship.
8. Liu Zheng, Sun Tan, Sheng Chun Lei, Liu Xiu min, National Science Library, China
Building semantic knowledge organization systems for interdisciplinary research: An example of wetland remote sensing [presentation ppt]
Abstract: The content of interdisciplinary research often scatters in various resources and difficult to describe systematically, comprehensively and accurately by using specialized or general Knowledge Organization System (KOS). Traditional knowledge organization tools show poor performance in tackling this problem. Classification can represent information in broad fields, but is too redundant to be precise. Topical representation can be useful for fine-grained matching, but the meaning of concept in a thesaurus or ontology tend to be too narrow for multi-disciplinary topics. In addition, it is time consuming and costly to build a completely new KOS. In this paper, we developed a set of procedures and methods to construct a semantic KOS, which contains a set of concepts and relationships and was built by using existing KOS, in support of interdisciplinary scientific research. We used the methods of domain expert participation and semantic matching and chosen wetland remote sensing as target topic field. We discuss how we solved problems involved in reusing existing KOS, selecting concepts and integrating semantic relations, represented knowledge with multi-dimensions, studied automatic construction rules and algorithms, and finally developed a semantic KOS for wetland remote sensing research. Each concept in this semantic KOS was assigned one of six specific semantic types, including satellite and sensor, aviation remote sensing and sensor, remote sensing data and products, remote sensing characteristics and application as well as data processing and method.
9. Sophy Shu-Jiun Chen, Academia Sinica, Taiwan
Art & Architecture Thesaurus (AAT) Extensions for Cultural-Specific Concepts: Methods, Applications & Issues [presentation ppt]
Abstract: This report first presents the methodologies of developing Chinese-Language Art & Architecture Thesaurus (AAT) which are based on the collaboration with the Getty Research Institute. The research modules of mapping, translating and adjusting process both from English to Chinese and from Chinese to English will be discussed. Then, how a multilingual micro-thesaurus has supported Chinese culture and linked data for digital libraries will be demonstrated. Finally, a number of issues including term mapping types, compound terms, loanwords and Disambiguation during the whole lifecycle of construction will be discussed further.
10. Xia Lin, Drexel University, USA
Facet-based Mapping Interfaces for KOS vocabularies [presentation ppt]
Abstract: Mapping concepts between two Knowledge Organization Systems (KOS) vocabularies has been a major challenge for effective use of KOS in practical systems. We are currently developing a pattern-based technique for such mapping. The technique works well on facet-based vocabularies with which concepts can be decomposed through facets and elemental concepts. The strings of the elemental concepts then can be mapped to concepts of the other KOS through search engines or semantic mapping. To demonstrate this process, two facet interfaces will be presented and compared in this presentation. The interfaces will map Getty’s AAT thesaurus terms to ARTstor’s indexing terms in assisting user’s searching and browsing. First, the user’s free-text query will be mapped to multiple concepts arranged by the facets. The user then can select one or more concepts from the displays to map them to ARTstor indexing terms. Finally, the user can select the indexing terms to complete their searching or browsing. Through this mapping process, the interfaces help the user convert their initial queries to precise and KOS-based queries for the purpose of high-precision searching and browsing. The central idea we are advancing through the interfaces is how to combine the user’s selection process with the computer’s matching process to provide an effective mapping between two KOS vocabularies.