Greenstone is a comprehensive system for constructing and presenting collections of thousands or millions of documents, including text, images, audio and video.

A typical digital library built with Greenstone will contain many collections, individually organized—though they bear a strong family resemblance. Easily maintained, collections can be augmented and rebuilt automatically.

There are several ways to find information in most Greenstone collections. For example, you can search for particular words that appear in the text, or within a section of a document. You can browse documents by title: just click on a book to read it. You can browse documents by subject. Subjects are represented by bookshelves: just click on a bookshelf to look at the books. Where appropriate, documents come complete with a table of contents: you can click on a chapter or subsection to open it, expand the full table of contents, or expand the full document into your browser window (useful for printing). The New Zealand Digital Library website ( nzdl.org ) provides numerous example collections.

On the front page of each collection is a statement of its purpose and coverage, and an explanation of how the collection is organized. Most collections can be accessed by both searching and browsing. When searching, the Greenstone software looks through the entire text of all documents in the collection (this is called “full-text search”). In most collections the user can choose between indexes built from different parts of the documents. Some collections have an index of full documents, an index of paragraphs, and an index of titles, each of which can be searched for particular words or phrases. Using these you can find all documents that contain a particular set of words (the words may be scattered far and wide throughout the document), or all paragraphs that contain the set of words (which must all appear in the same paragraph), or all documents whose titles contain the words (the words must all appear in the document's title). There might be other indexes, perhaps an index of sections, and an index of section headings. Browsing involves lists that the user can examine: lists of authors, lists of titles, lists of dates, hierarchical classification structures, and so on. Different collections offer different browsing facilities.

Greenstone constructs full-text indexes from the document text—that is, indexes that enable searching on any words in the full text of the document. Indexes can be searched for particular words, combinations of words, or phrases, and results are ordered according to how relevant they are to the query.

In most collections, descriptive data such as author, title, date, keywords, and so on, is associated with each document. This information is called metadata. Many document collections also contain full-text indexes of certain kinds of metadata. For example, many collections have a searchable index of document titles.

Users can browse interactively around lists, and hierarchical structures, that are generated from the metadata that is associated with each document in the collection. Metadata forms the raw material for browsing. It must be provided explicitly or be derivable automatically from the documents themselves. Different collections offer different searching and browsing facilities. Indexes for both searching and browsing are constructed during a “building” process, according to information in a collection configuration file.

Greenstone creates all index structures automatically from the documents and suppporting files: nothing is done manually. If new documents in the same format become available, they can be merged into the collection automatically. Indeed, for many collections this is done by processes that awake regularly, scout for new material, and rebuild the indexes—all without manual intervention.

Source documents come in a variety of formats, and are converted into a standard XML form for indexing by “plugins.” Plugins distributed with Greenstone process plain text, HTML, WORD and PDF documents, and Usenet and E-mail messages. New ones can be written for different document types (to do this you need to study the Greenstone Digital Library Developer's Guide). To build browsing structures from metadata, an analogous scheme of “classifiers” is used. These create browsing indexes of various kinds: scrollable lists, alphabetic selectors, dates, and arbitrary hierarchies. Again, Greenstone programmers can create new browsing structures.

Collections can contain text, pictures, audio and video. Non-textual material is either linked into the textual documents or accompanied by textual descriptions (such as figure captions) to allow full-text searching and browsing.

Unicode, which is a standard scheme for representing the character sets used in the world's languages, is used throughout Greenstone. This allows any language to be processed and displayed in a consistent manner. Collections have been built containing Arabic, Chinese, English, French, M 0Å 1ori and Spanish. Multilingual collections embody automatic language recognition, and the interface is available in all the above languages (and more).

Collections are accessed over the Internet or published, in precisely the same form, on a self-installing Windows CD-ROM. Compression is used to compact the text and indexes. A Corba protocol supports distributed collections and graphical query interfaces.

The New Zealand Digital Library ( nzdl.org ) provides many example collections, including historical documents, humanitarian and development information, technical reports and bibliographies, literary works, and magazines.

Being open source, Greenstone is readily extensible, and benefits from the inclusion of Gnu-licensed modules for full-text retrieval, database management, and text extraction from proprietary document formats. Only through international cooperative efforts will digital library software become sufficiently comprehensive to meet the world's needs with the richness and flexibility that users deserve.