Greenstone tutorial exercise

Back to wiki
Back to index
Sample files: marc.zip
Devised for Greenstone version: 2.60|3.06
Modified for Greenstone version: 2.87|3.08

Bibliographic collection

This exercise looks at using fielded searching in a collection. Fielded searching is best used for metadata rich collections. Here we use bibliographic data in MARC format.

  1. Start a new collection called Papers Bibliography which will contain a collection of example MARC records of the working papers published at the Computer Science Department, Waikato University. Enter the requested information and base it on -- New Collection --.

  1. In the Gather panel, open the sample_files → marc folder, drag CMSwp-all.marc into the right-hand pane and drop it there. A popup window asks whether you want to add MARCPlugin to the collection to process this file. Click <Add Plugin>, because this plugin will be needed to process the MARC records.

  1. Now select Browsing Classifiers within the Design panel and remove the default classifier for Source metadata.

  1. In the Search Indexes section, remove the ex.Source index. In this collection all records are from the same file, so ex.Source metadata, which is set to the filename, is not particularly interesting or useful.

  1. Switch to the Create panel, build the collection, and preview it. Browse through the Titles and view a record or two. Try searching—for example, find items that include graphics.

  1. Back in the Librarian Interface, go to the Browsing Classifiers section of the Design panel. Select AZCompactList from the Select classifier to add drop down menu, and click <Add Classifier...>. In the popup window, select dc.Subject and Keywords as the metadata item. Click <OK>.

    AZCompactList is like List, except that terms that appear multiple times in the hierarchy are automatically grouped together and a new node, shown as a bookshelf icon, is formed.

  1. Build the collection and preview the result.

Using fielded searching

  1. Now let's look at fielded searching. In the browser, go to the PREFERENCES page. You will notice that there is a Query style: option which enables you to switch between "normal" and "fielded" search. Change to fielded search now, press the set preferences, and click on the Search button to go back to the Search page. The search form has changed to a fielded form.

  1. You can specify which search form types are available for a particular collection, and which one is the default, using the searchType format statement. In the Format panel, select Format Features from the left-hand list. Select the searchType format statement from the list of assigned formats, and set the contents to form. This will make only fielded searching available for this collection.

    Search type options include form and plain. You can specify one or both separated by a comma. If both are specified, the first one is used as the default: this is the one that the user will see when they first enter the collection.

  1. Preview the collection again. Notice that the collection's home page no longer includes a query box. (This is because the search form is too big to fit here nicely.) To search, you have to click Search in the navigation bar. Note that the PREFERENCES page has changed so that the "normal" query style is no longer offered.

  1. Look at the search form in the collection. There are two fields that can be searched: text and titles. Add some more fields to search on by going back to the Librarian Interface.

  1. In the Design panel, go to the Search Indexes section. Add a new index based on dc.Subject and Keywords by clicking <New Index>, selecting dc.Subject and Keywords in the list of metadata, and clicking <Add Index>.

  1. Rebuild the collection and preview the results. Notice the extra field in the ... in field drop-down menus in the search form. You can do quite complicated queries by searching for words in different fields at the same time.

  1. To change the text that is displayed in the drop-down menus of the search form, you would go to the Search section of the Format panel. Here you can change the display text for the indexes.

Exploding the database

  1. Go to the Enrich panel and try to see the metadata. It doesn't appear! This is because the metadata is associated with records inside the file, not the file itself.

    Metadata file types, such as MARC, CDS/ISIS, BibTex etc. can be imported into Greenstone but their metadata cannot be viewed in the Librarian Interface. To edit any metadata you need to go back to the program that created the file.

    Greenstone provides a way of exploding a metadata database so that each record appears as an individual document, with viewable and editable metadata. This process is irreversible: once this step has been done, the database is deleted and can no longer be used in its original program.

  1. In the Gather panel, you may notice that the MARC database has a different coloured icon to other files. A metadata database that can be exploded will be displayed with this green icon. Right-click on the file and choose Explode Metadata Database from the menu. A new window opens, containing options for the exploding process. A description of each option can be obtained by hovering the mouse over the option.

    If it's not already on, turn on the metadata_set option by checking its box. This option indicates which metadata set to explode the metadata into. The default set is the "Exploded Metadata Set"—a metadata set which initially has no elements in it, but will receive a new element for each metadata field retrieved from the database.

  1. Click <Explode> to start the exploding process. This may take a short while, depending on the size of the database.

  1. Once exploding has finished, the MARC database file will have been deleted, and three folders created in its place. These folders contain an empty file for each record in the original database. The metadata for these records can be viewed and edited by switching to the Enrich panel.

  1. Because the MARC file is no longer present, and the collection contains empty (.nul) files, we need to change the list of plugins. In the Document Plugins section of the Design panel, remove MARCPlugin.

  1. Rebuild and preview the collection. You will notice that the Subjects classifier is empty, searching no longer returns any results, and the document display is useless.

    Although the Titles classifier was built on ex.Title, it still displays the correct titles, but in the Enrich panel you can see the ex.Title metadata are actually the filenames rather than titles of the MARC records. This is because the default VList format uses the exp.Title metadata. In the Format Features section of the Format panel, select VList in the list of assigned format statements. The format statement looks like:

    <td valign="top">[link][icon][/link]</td>
    <td valign="top">[ex.srclink]{Or}{[ex.thumbicon],[ex.srcicon]}[ex./srclink]</td>
    <td valign="top">[highlight]
    {Or}{[dc.Title],[exp.Title],[ex.Title],Untitled}
    [/highlight]{If}{[ex.Source],<br><i>([ex.Source])</i>}</td>

    Since there is no dc.Title metadata and because exp.Title comes before ex.Title, the exploded titles will be displayed.

Reformatting the collection to use the exploded metadata

The collection previously used extracted (ex.) metadata, but now it uses exploded (exp.) metadata. The Subjects classifier and search indexes were built on ex metadata, which is why they no longer work properly.

There is also no longer any text in the documents. Previously, MARCPlugin stored the raw record as the "text" of each record. Now that the metadata is in the Librarian Interface, there is no longer the concept of raw record, and so there is no text.

We need to modify the collection design to take note of these changes.

  1. In the Search Indexes section, change the Title index to use exp.Title: select the Title index in the Assigned Indexes list and click <Edit Index>. Deselect dc.Title and ex.Title in the list of metadata, and select exp.Title. Click <Replace Index>.

  1. Remove the dc.Subject and Keywords index by selecting it in the Assigned Indexes list and clicking <Remove Index>. Add an index on exp.Subject: click <New Index>, select exp.Subject in the metadata list, and click <Add Index>.

  1. The text index is no longer any use, so remove that index too.

  1. To enable combined searching across all indexes at once, click <New Index>, tick the Add combined searching over all assigned indexes (allfields) checkbox, and click <Add Index>. Move this to the top of the list using the <Move Up> button, so that it appears first in the drop down list. Click <Set Default Index> on the right so that it becomes the default field for searching.

  1. To explicitly use the exp.Title metadata, in the Browsing Classifiers section, change the dc.Title;ex.Title List to use exp.Title metadata. Double click the dc.Title;ex.Title List in the Assigned Classifiers list, and change the metadata option to use exp.Title. Click <OK>. Do the same thing for the Subject AZCompactList, changing dc.Subject and Keywords to exp.Subject.

  1. Rebuild and preview the collection. The classifiers should be back to normal and searching should now work.

  1. In the Format Features section of the Format panel, select VList in the list of assigned format statements.

    The resulting format statement looks like:

    <td valign="top">[link][icon][/link]</td>
    <td valign="top">[highlight]
    {Or}{[exp.Title],[ex.Title],Untitled}
    [/highlight]</td>

  1. Clear the DocumentHeading format statement by selecting it in the list of assigned format statements and deleting the contents in the HTML Format String. The record Title will be displayed as part of the DocumentText format, so we don't need it here.

  1. Next, edit the DocumentText format statement. Delete the contents and replace it with the following (which can be copied from sample_files → marc → format_tweaks → document_text.txt).

    <table>
    <tr><td>Title:</td><td>[exp.Title]</td></tr>
    <tr><td>Subject:</td><td>[exp.Subject]</td></tr>
    <tr><td>Publisher:</td><td>[exp.Publisher]</td></tr>
    </table>

  1. The DETACH and NO HIGHLIGHTING buttons are not very useful for this collection, so lets get rid of them. Edit the DocumentButtons format statement to make it empty. Press the <Preview Collection> button to preview the collection and see how the document display has improved.


Copyright © 2005-2016 by the New Zealand Digital Library Project at the University of Waikato, New Zealand
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation; with no Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts. A copy of the license is included in the section entitled “GNU Free Documentation License.”