This document specifies the metadata used in the Lund university dissertation database.
The current Dissertation abstract service [1] at Lund university is based on free text search on a weakly fielded record model. By that statement I understand that it is possible to identify where in the HTML document used as the record syntax information of given type is found. It is thus possible to generate these records by HTML forms, and also to edit the metadata through this procedure.
Although the database could be said to be "fielded", it is impossible to search it that way. Also, the collection of records are in state of decay; the data has been created with different tools, under different operating systems with different character encodings. This, in turn, means that recall as well as precision has been suffering.
During the autumn 2001, a new prototype user interface with enhanced search system [2] (please note that the search system might not be operatational at all times). The work included an export facility, making it possible recover a fairly large proportion of the legacy database. The content model of the prototype service was documented using a XML dtd 3. This paper is the result of the discussions that followed, and articulates the modifications that we have made on the metadata to satisfy the functional requirements of the long overdue service. Further documentation exists in the form of a revised DTD [7] and a manually prepared record implementing it [8]
This is not an academic exercise, since the data model of a service is something we have to live with in this service, and also something we will have to explain to users, and also something we will have to pay for in terms of human labour.
A dissertation is a publication, and the basis for my model is basically a bibliographic record. However, it is in several ways more detailed, since a dissertation is connected to an event, the defense act. The defense act is an event which occurs at certain place and at a certain time, and it involves a special person, the external examiner employed by the faculty.
A dissertation may be a traditional monograph, but more often than not, a dissertation is a composite document. In this case it consists of an abstract, a summary which in itself is a sizable document, and a number of articles.
Considerations as those above gave me the following elements (this is not a complete list, since I have omitted some that are used internally, auto-generated, or are of less importance).
The top element of the DTD is 'dissertation'. The element has two attributes 'status' and fulltext (with values from the enumerated list: submitted, issued and failed) (accepting the values 'public', 'archived' and 'missing').
The lists below show the exact order in which the elements must occur in order to be valid according to the DTD. An element followed by '?' are optional or may occur exactly once, whereas if an element is followed by '*', the element may be included zero or more times. '|' implies a choice. For instance you may have either an arbitrary number of 'component-journal-article' or zero or one single 'component-blob'.
creator, supervisor, title, pages, language, availability, date-issued, defended, uncontrolled-terms?, controlled-terms?, controlled-umi-term, summary-in-english, summary-in-swedish?, funded-by*, component-blob?| (component-journal-article*| component-article-in-book*| component-report*| component-eprint*) rights-management
A creator is a person. In Scripta, a person has the following characteristics:
name, born?, electronic-address?, affiliation?
The meaning characteristics should be obvious. Date of birth is available and is widely used in bibliographic records. One may question if date of admittance to a post-graduate program is relevant, and the same is true for gender. 'address' is the electronic one. name has a content model of itself:
#PCDATA|(last,first)
that is, you may either enter a string or surname and given name. Affiliation is an organization, or possibly a division within an organization. Organizations are described as
name?, (name-hierarchy|department)?, home-page?, postal-address?
I.e., the record contains either a name-hierarchy or a department. Name hierarchy is of the form "faculty / department / sub-department" and applies to affiliations within Lund university. It will permit search for the exact match or sub-string, i.e., give me all dissertations within the faculty of science or give me all dissertations in Limnology (search for string and switch on complete field).
The availability element follows the GILS Z39.50 protocol [4] and is structured into the following:
available, medium, distributor
'medium' is the SAFARI document type of the thesis, i.e., text.thesis.doctoral. A distributor is an organization and described as for affiliation above.
Available has the following content model:
linkage?, issn?, series-title?, isbn?, coden?
The linkage is the URL of the thesis cover page in our archive, if available from that source. The structuring of availability is dependent on the GILS tag-set [4].
The 'defended' element is unique to dissertations:
date, time, location, external-examiner
I.e., when, where and who does the dirty job. opponent is a person, and has thus the same content model as creator.
The dissertations are indexed using uncontrolled-terms and controlled-terms. Each of these may include an arbitrary number of uterm and cterm respectively. The current prototype database is using CERIF as the controlled subject vocabulary [5] (available in the tree format [6]. The version of CERIF has some shortcomings as regards its structure, and its worst is medicine.
In addition to CERIF, we also need a UMI subject [9-10], in order to export data to UMI's database. This subject is stored in the element 'controlled-umi-term', which is mandatory and not repeatable.
The funded-by element names the sponsoring organisation(s) and is repeatable. Like affiliation and distributor, it is an organisation.
A valid dissertation description must contain a summary-in-english and may include a summary-in-swedish. It has been regarded as a functional requirement that these should be give a minimum of typographical adornment. The DTD defines a typographic content model containing the following tags for this purpose:
ol| ul| p| sub| sup| strong| em
They are borrowed directly from HTML and means ordered list, un-ordered list, paragraph, sub-script super-script, bold and emphasis, respectively. The tag 'li', list item, may be used in both list types.
The typographic content model applies to 'title' and 'cross-reference-blob' as well, the former that I felt that our database already contain titles with super- and sub-scripts. Also, emphasis is used for species names in paleontology, biology and medicine. Actually, contrary to the DTD, the database contain quite a few records that use emphasis on the uncontrolled terms for species names. (All records are well-formed XML, only a minority are valid).
The content model permits two forms of cross-references, or component publications. The way this concept has been introduced into the DTD is borrowed directly from GILS[4], but it is subject to revision. The cross-reference-blob is the home for legacy cross references, there is no way one can correct their format automatically.
The dissertations components are expected to be of three kinds article in book, journal article and reports, where the latter are expected to contain publications of various shades of grey. Patents goes into this category as well. These three correspond to the following three elements: component-journal-article, component-article-in-book, component-report and component-eprint. All of these elements have the following common substructure in common:
creator+,
date?,
title,
appeared-in-title?,
volume?,
issue?,
pages?,
isbn?,
issn?,
coden?,
editor*,
part-of-series-title?,
publisher*,
locator*
Where the semantics of the elements are defined as follows.
| creator | The name of an author. A person as for the other persons |
| book-title | For an article that is part of a book, the title of the book |
| place-published | The place of publication of a book. |
| date | The date of publication, yyyy. |
| editor | For an article that is part of a book, the name of an editor of the book. |
| publisher | The publisher of a book or report series. |
| appeared-in-title | For articles, the title of the journal or book the work appeared in. |
| issue | The number of the journal issue (or, if applicable, of the report itself. |
| volume | Journal volume number. |
| part-of-series-title | In case of an article or a report, the name of the book or report. |
| title | The name of the component work. Typically a title of an article. |
| locator | For any component medium a pointer to an electronic version of the object. |
Each of these elements accepts the attribute status taking either of the values 'manuscript', 'submitted', 'inpress' and 'published'.
Finally, the DTD includes an element 'rights-management', meant to be presented as an acknowledgment to a publisher.