Internationalized Back-of-the-Book Indexes for XSL Formatting Objects

The XSL recommendation and its commercial implementations have finally reached a point of maturity that allows for their use in generating production-quality printed technical documents. However, the XSL specification does not provide any built-in features specific to the production of back-of-the-book indexes. In addition, the current commercial tools do not provide any proprietary features for indexing comparable to those that can be found in existing page composition systems such as Arbortext's Epic Publisher and XyEnterprise XPP. Thus, if XSL-FO is to be used to produce documents that have back-of-the-book indexes, the style sheet author must implement all the processing necessary to create the index. This task is challenging enough for Latin languages. For indexes in other languages, the task is further complicated by the need to support locale-specific collation schemes and index groupings. Defining collation and grouping rules for languages such as Traditional and Simplified Chines, that use tens of thousands of unique characters, further complicates the task. This paper describes in detail the system developed by the authors to satisfy the challenge of producing back-of-the-book indexes using XSL-FO for a number of Asian and Middle-Eastern languages, including Thai, Japanese, Korean, Arabic, Hebrew, and Chinese (Traditional and Simplified). The solution developed includes the following key components: • XSL business logic for processing the index entries in order to generate the index pages themselves • Java libraries that support index item grouping and sorting for arbitrary national languages and locales, configured through an XML document by which the index groups and sorting rules are easily specified • a PDF post processing step that removes duplicate page numbers from the index pages produced by the XSL-FO process. The solution developed is completely generic and can be easily adapted to any typical technical documentation document type that uses embedded index entry markup (e.g., any Docbook-like document type). The configuration of the indexes is through relatively simple XML documents that define the index groups, sorting and grouping strategies to use and, if necessary, locale-specific collation orders for index entry and group sorting. The paper describes how the system takes Rendered by www.RenderX.com full advantage of Java's built-in internationalization support to simplify both the initial development task and the definition of custom index configurations. The paper specifically discusses the indexing challenges posed by the national languages involved and how the XSL and Java code developed meets those challenges. Also discusses the general issues involved in generating indexes with XSL-FO, including issues of locale-specific sorting, grouping using the Munchian Method, and use of Java's built-in support for locale-specific collation. Internationalized Back-of-the-Book Indexes for XSL Formatting Objects Table of