BIBFRAME核心类演变分析

Evolution of BIBFRAME Core Classes

BIBFRAME 2.0草案提出核心类由创作作品、实例、规范和注释四个变为创作作品、实例和单件三个。从关联数据角度分析,核心类演变的原因主要有三个方面:关联数据通过资源URI确认身份的唯一性,不需要规范检索点;关联数据词表不需要重新定义所有类和属性,可以直接使用已有词表,包括Web注释数据模型和其他成熟的词表或本体;关联数据以三元组揭示资源,可以跨越MARC书目格式和馆藏格式的界限。

Bibliographic Framework, a linked data （LD） initiated by the Library of Congress （LC） for the replacement of MARC format, started in 2011. BIBFRAME proposed model was issued at the end of 2012, with four main/core classes Creative work, Instance, Authority and Annotation. Creative work class and Instance class are considered a simplification of FRBR＇s WEMI model. Authority class continues the tradition of authority control in librarianship. Annotation class is a mixture of elements other than bibliographies and authorities, including emerging internet things such as cover arts and online reviews, and library holdings information. BIBFRAME vocabulary was released at the beginning of 2013 and basically established in 2014. LC and other implementers experimented it for more than one year. Based on discussion in experiments and suggestions from experts, LC had issued five revision proposals since June 2015 and announced seven BIBFRAME 2.0 drafts in October 2015. According to the proposals and drafts, BIBFRAME vocabulary will evolve in various aspects with the most noticeable adjustment of core classes, i.e. cancel of Authority and Annotation and introduction of Item. The paper attempts to analyse the significant improvements in BIBFRAME＇s LD practice as showed in the evolution of its core classes, and hopes that it could help catalogers and other people with a better understanding of LD and its development in bibliographic application. 1） Traditional authority control in librarianship is mainly dependent on the authorized form of names, i.e. authorized access points. Due to the diversity of languages, it＇s difficult to agree on the authorized form between countries. Application of authority control outside libraries is even more difficult. Concepts and things are identified by URIs in LD, without relying on the authorized access points. URI can be compatible with authorized form in ＂Preferred label＂ meanwhile. Cancel of Authority class makes clear that name authority is only a property of things in the context of LD. 2） Reuse of mature vocabularies is a generally accepted best practice in LD. At the beginning of BIBFRAME development, LC used single namespace and concerned about the persistence of external namespaces. As Web Annotation Data Model will be a W3C Recommendation, progress of RDF vocabularies in recent one or two years makes LC change its practice. Cancel of Annotation class confirms vocabulary reuse. BIBFRAME 2.0 drafts also use rdf：/rdfs： namespaces and show that BIBFRAME will focus on the core bibliographic data. For fully converting of MARC data and accurate identification of resource, BIBFRAME will inevitably reuse partly other vocabularies, such as EBUCore/PBCore, for complex properties of special resources. 3） In Bibliographic Framework Initiative General Plan, holdings are among the requirements to support bib- liographic description. Holding is a data type used for ILS＇s circulation module. Newly added Item class is a redefinition of holding-related sub-classes of original Annotation class. Unlike ILS which bases on biblio- graphic and holding/item records, LD uses triple to identify resource and needs no records. Simple and compound items in BIBFRAME 2.0 leap over the clear distinction of bibliographic and holdings format/re- cord, and can reveal item information in a more flexible, lightweight and accurate way. When initiating BIBFRAME modeling, LC＇s major focus is ＂to translate the MARC 21 format to a Linked Data （LD） model＂. That is the reason why BIBFRAME vocabulary retains a large number of traces of MARC format. As the experiment of BIBFRAME moves from ＂transformation＂ to ＂original cataloging＂, describing resources directly in BIBFRAME becomes the focus. Evolution of BIBFRAME 2.0 core classes shows that understanding and application of LD in librarianship is deepgoing, and LD itself is also continuing to mature. 13 refs.