Into the Depths of Data. Methods of Subject Specific Content Retrieval

Into the Depths of Data. Methods of Subject Specific
Content Retrieval

Kurt
Gärtner

University of Trier
gaertnek@mailer.uni-marburg.de

Gisela
Minn

University of Trier
minn@uni-trier.de

Andrea
Rapp

University of Trier
rappand@uni-trier.de

Martin
Raspe

University of Trier
raspe@uni-trier.de

Ruth
Christmann

University of Trier
christma@uni-trier.de

Thomas
Schares

University of Trier
schares@uni-trier.de

2002

University of Tübingen

Tübingen

ALLC/ACH 2002

editor

Harald
Fuchs

encoder

Sara
A.
Schmidt

In April 1998, the Competence Centre for Electronic
Retrieval and Publishing Techniques in the Humanities was founded
at the University of Trier. The use of international hard- and software
independent standards as SGML/XML is one of the main targets of the
Competence Centre in dealing with full-text digitization especially of
critical editions, dictionaries, and important reference works. Information
scientists and humanists from various disciplines are working closely
together in order to guarantee that the electronic resources developed at
the Centre meet with scientific requirements. Furthermore, the team aims at
complex and powerful retrieval mechanisms that can be handled easily by a
consistently user-oriented design of Graphical User Interfaces. An important
overall feature that has often been ignored by people working in the field
of digitization but is characteristic for the research done at the
Competence Centre is the close linking of software development to the
scholarly background of the material.
Examples for the development of user-oriented software in different projects
as well as for the embedding of the activities of the Competence Centre into
research done by universities and the German academies of sciences shall be
given in the following three papers on (A) the Rhine-Meuse Net, (B) the WIRE
project, and (C) the digitization of the Deutsche
Wörterbuch - a history, an art history, and a German language
and literature project.
(A) The conception of the so-called Rhine-Meuse Net originated from the
activities of a Collaborative Research Centre (= SFB 235) having examined
the history of a European core area from the Ancient World to the 19th
century. For more than 12 years, a large amount of valuable data has been
accumulated in multiple document types and formats. However, not all the
material was published, although, in many cases, even the unpublished
material is of high interest to researchers in and outside the context of
the SFB. Therefore, the existing data will now be encoded in order to ensure
its longevity and at the same time be entered into a database. Thus it will
be possible to use these data even though the funding of the SFB by the
Deutsche Forschungsgemeinschaft (= DFG) is due to cease in 2002.
(B) In contrast to the Rhine-Meuse Net dealing with material already
existing, WIRE, the Word and Image Retrieval Environment, is primarily intended as a tool for
scholars that need some support in building new (digital) collections of
scientifically relevant texts and images. The internet-based system allows
for an integration of texts, structured data, images, and bibliographies
into a relational database. As WIRE can be configured according to specific
needs, it does not only support the use by individual scholars but is also
well apt at being used by teams of scholars working together on a particular
object of research. Since various retrieval functions are implemented, WIRE
is not only useful for scholars who build new collections but also for those
who only want to browse through collections built by their colleagues.
(C) The retrodigitization of the Deutsche Wörterbuch
by Jacob and Wilhelm Grimm has to be seen in the broader context of
dictionary making at the University of Trier. When work on a new Middle High
German dictionary was started in 1994, lexicographers wished to have access
to as many electronic texts and dictionaries as possible. However, to fully
exploit the advantages of an electronic dictionary, one does not only need a
fairly thorough markup of the entries but also a highly comfortable way to
present the dictionary on screen and thus make it readable - just imagine
that several entries of the Deutsche Wörterbuch
cover more than 300 columns in print! The demonstration of the CD-ROM
prototype of the Deutsche Wörterbuch might serve as
a good example for how in-depth retrieval carried out thoroughly contributes
to the development of software that allows accessing the dictionary data in
new ways.
It will be very interesting to see how new possibilities to access data of
various provenance and of multiple kinds will lead to new questions, new
methods, and new insights into the digitally edited source material.

Title A: The Information and Reference Network for the History of the
Rhine-Meuse Area. An Area-Oriented Subject Information System for the
Humanities
Dr. Gisela Minn
Dr. Andrea Rapp

1. General and Institutional Preconditions
Apart from the parameter "time", the parameter "area" has in the
past few years received increased attention as a fundamental
category of human existence. Particularly regions as
middle-sized units of area have established themselves in a
multitude of disciplines as ideal units for investigation. In
the Rhine-Meuse Net, the regional area is made use of as a
central access and ordering category for the integration of
research results that are far apart with regard to time and
differ in document type, methods, and topic. The international
research compound of the Collaborative Research Centre "Between
the Meuse and the Rhine. Connections, Encounters, and Conflicts
in a European Core Area from the Ancient World to the 19th
Century" (SFB 235) has acquired a large amount of valuable and,
with regard to document types, very heterogeneous data, that are
not only concerned with a common area of investigation but are
also closely connected with regard to content.
This complex amount of data forms the nucleus of a projected
database serving as a reference system for European regional
history. The project is being funded by the Deutsche
Forschungsgemeinschaft (DFG) since 1st November 2001. Apart from
the historical field with all its specialist research interests,
there are involved related disciplines such as art history,
archaeology, history of law, and history of German and Roman
languages; they all partake in the research compound, as well as
various national and international, university and
non-university cooperation partners.
Therefore the project aims firstly to take into account the
changed needs for information of a growing international
research community and secondly to lay the grounds for European
research in history beyond the borders of nation-states. For
this the network is particularly apt, as it opens up a European
core area at the intersection between Western and Middle Europe
from ancient times up to the present, and it will present the
results of international researchcollaboration. The long-term
data-conservation and its platform-independent use is ensured by
a consistent application of international standards on the basis
of SGML/XML.

2. Content-Related Principles of the Network
The realization of the network starts at two core units: Firstly,
the annotated bibliography of the whole publication output of
the SFB (about 900 nos.) will be edited, including all the
unpublished dissertations and theses which document the whole
scope of research. Due to the area-oriented interest of the SFB,
cartographical methods and techniques of representation belong
to the most important research procedures. Thus secondly, an
electronic archive of maps was built (of about 500 items) that
will be linked to the bibliography.
By these two core units that are representative for the whole
scope of the network, thesauri of places, persons, and subjects
will be accumulated and structured hierarchically for an
in-depth disclosure of the data. They form the basic framework
for a further indexing of the data and will be extended to a
dynamic research tool that will become more extensive and
complex with the integration of each new reference unit. A
sophisticated system of indexes and metadata will guarantee the
linking of these units.

3. Variety of Document Types
The document types representing the cultural heritage as well as
the results of scientific research in digital form are very
heterogeneous: texts, maps, pictures, plans, images, tables,
archival finding-aids and repository guides, indices,
bibliographies etc. At the same time, these document types are
very closely related as regards content in a very complex and
multidirectional manner. In the Humanities especially,
far-reaching methodical and content-related impulses are to be
expected by an explicit representation of these relations.
Moreover, the general approach requires interdisciplinary and
comparative studies, new access to digital resources, and the
development of cartographic methods for analyzation and
documentation. Therefore, we aim at a concatenation, retrieval
and integration of these digital reference-units of different
document types in a reference compound. The following document
types form the database of the network and have to be opened up
and interlinked:
Units of information referring to area and region such
as local registers and catalogues, complex place lexica,
single maps and series of maps, annotated atlasses that
combine maps, place catalogues, and commentaries.
Units of information referring to persons and
institutions such as registers and catalogues of
persons, prosopographies and biograms of persons,
catalogues, lexica, tables, and lists of institutions.
Units of information that combine information on texts
and pictures such as text or picture catalogues,
visualizations, and reconstructions.
Units of information that represent sources, archival
finding-aids, and instruments for the documentation of
research such as special bibliographies, region-related
source editions of different genres, and repository
guides, literature and review service, documentations of
research.

Therefore we provided for the following ways of access by
thesauri: access by place (in addition by visual representations
such as two- or three-dimensional maps), access by time, access
by person, access by topic, access by object via document types
(e.g. only maps, only sources, etc.), access by funding
organizations (respective research institution).

4. Methods, Technical Bases
Due to the complexity of the structures and the implicit
relations characteristic for the Humanities, the construction of
such networks cannot be carried out by automatic means only but
has to be completed and supervised by human researchers.
Therefore it is all the more important to develop mechanisms
with the help of standards that support the construction of a
complex structure and corresponding retrieval mechanisms
effectively. Moreover, these mechanisms have to be well
documented and safely stored for further research in times of
rapid technical change. Variable, differentiated, and efficient
strategies for searches and visualizations have to be created
for convenient use.
Due to different structures of document types brought together in
the network, existing DTD schemes have to be checked as to their
usability, varied and expanded, and new DTD schemes have to be
developed for document types not already on hand in an
SGML-compliant format. These schemes have to be applicable to
different research projects as well. The open conception of the
network resulting from this is the precondition for a transfer
of these structures and methods to other information and
reference networks.

5. Comparisons and Prospects
The information and reference network is open to cooperation
projects with university and non-university institutions
offering further information that goes beyond the scope of
research of the SFB. Special emphasis is laid on the integration
of libraries and archives. For example, the SFB bibliography as
core unit is linked to the OPAC of Trier University Library. The
integration of archival finding-aids beginning with the
finding-aids of the municipal archive Worms may serve as an
example for cooperation with other archives. Furthermore,
cooperations with scholars from neighbouring countries have been
established which focus on common region-related aspects and
methods respectively and, by common use of the information and
reference network, should be long-lasting.
In some regards, the Rhine-Meuse Net was inspired by the project
"The Valley of the Shadow. Two Communities in the American Civil
War", that was carried out at the University of
Virginia/Charlottesville ().
Especially the regional aspect as well as the variety of
document types offered are comparable to the content and
structure of the Rhine-Meuse Net. However, a significant
difference can be seen in the variety of topics of the documents
worked on in the Rhine-Meuse Net and in the in-depth retrieval
and thorough interlinking of these materials.

6. Literature

Franz
Irsigler

Raumkonzepte in der historischen
Forschung

Zwischen Gallia und Germania, Frankreich und
Deutschland. Konstanz und Wandel raumbestimmender Kräfte
Trierer Historische Forschungen

Hrsg. von

Alfred
Heit

12

1987
11-27

Andrea
Rapp

Die elektronische Publikation, Erschließung und
Vernetzung des Trierer Korpus mittelfränkischer Urkunden des
14. Jahrhunderts

Jahrbuch für Computerphilologie

Hrsg. von

Georg
Braungart

Karl
Eibl

Fotis
Jannidis

Paderborn

2000
147-161

online in: Jahrbuch für Computerphilologie

RMnet

Title B: WIRE -- An Instrument for Collecting Visual and Textual Data
Martin Raspe

WIRE (“Word & Image Retrieval
Environment”) is an integrated environment for
collecting scientific resources, particularly in the field of
art history. It is being developed at Trier University and is
specifically taylored to meet the methodologic needs of the
discipline. Image data, bibliographic entries, source texts (and
optionally other structured data) are stored in a single
relational database, along with an unrestricted number of
descriptive texts which may contain individual formatting. The
whole material can be searched conventionally; in addition, the
content is accessible through a flexible, hierarchically
organized, multi-lingual keyword system. Data can be entered as
well as queried via Internet from different places at a time.
The program is based on widely-spread software components
(Microsoft Word plus a web browser) and is easy to learn.

1. The Problem
Not many art historians use database systems in their research
projects. This fact is partly explained by the general
reluctance of traditional scholars towards new technologies; on
the other hand, available database software doesn't lend itself
easily to the specific tasks of this discipline. The sources
consist of images and various types of historic texts which do
not fit easily into predefined categories or structures;
accordingly, the methodology of art history is mainly based on
associative rather than standardized procedures.
Consequently, the incoming data either has to be trimmed down to
fit in schematic entry forms, or the resulting overhead will
grow so complex that the efforts to manage the database soon
overshadow its practical use. Moreover, in the course of every
research project questions tend to come up that have not been
thought of when the database structure was first conceived.

2. The Idea
Thus the staff of the History of Art department at Trier
University asked for a program that manages text along with
images, is equally suited for the needs of teaching and
research, supports team cooperation at different places and can
be understood by scholars who have hardly any computing
experience beyond word processing. Departing from this
specification I started to look around for software, but it soon
turned out that existing programs - most of them authoring
systems - are much too complex and would strain the finances of
a small department. This situation led to the idea of creating
such a program myself. Since I am an art historian and have some
experience in computer programming, the task seemed feasible; as
an associated member of the "Competence Centre", I get all the
practical support I need.

3. The Concept
My intention was to create a kind of working tool for art
historians that supports the collecting of visual and textual
data. It soon became clear that the program had to focus on
three major types of material, i.e. images
of works of art, original source
texts and bibliographic
references, all three of which should be accessed
through textual descriptions. In addition it should be possible
not only to collect the material,
but also to arrange it, to comment on it and to add independent
scientific texts without restrictions. On top of that, people at
different places should be able to work simultaneously with the
same database.
WIRE tries to meet these requirements in a simple and robust, but
flexible way. It tries to combine two different types of data, a
collection of unstructured scientific
texts and a structured database
system that can be queried with exact criteria. The
texts that describe, summarize or comment the source material
form the scientific backbone of the database. Links can be
created from these texts to any of the source documents and
between them, so that the user is guided from one document to
other pertinent subjects. All texts can be searched fully, but
to retrieve the content intelligently a flexible keyword system
based on thesaurus lists is used. Each text may be associated
with any number of keywords; single keywords as well as entire
thesauri can be modified and added at any time without touching
the main data tables. The keywords can be structured
hierarchically, so that a search for "Tuscany" will also return
those entries which were only associated with the keywords
"Florence" or "Pisa."
A main goal is ease of use: All entry, modification and data
maintenance is done from Microsoft
Word, while the material may be searched and viewed
through a web browser. Data collections can be accessed locally,
but also via Internet from all over the world. Accordingly, WIRE
has a multilingual user interface (currently you can switch
between German and English; French, Italian and Latin are
planned). WIRE is not suited for highly specific databases with
many fields which require complex query strategies; in order to
achieve simplicity and flexibility some compromises are made.
Anyway, the collected material can be exported to other
databases using standard SQL commands; exporting into
standardized XML format is planned, thereby ensuring the future
value of the collected data.

4. The Realization

Database design
Every project that is realized in WIRE is an independent,
internet-accessible database that contains the complete data
(except for the image files which are stored in separate
directories). The three categories of source material and
the accompanying texts are stored in predefined tables
(which can be customized later). The bibliographic table has
the characteristic fields, while the image table contains
filenames and short identification tags. Each keyword list
is stored in a separate table; one of its fields denotes the
position of the entry in the hierarchy. To guarantee speed
and consistency, all links between documents and keywords
are stored in one heavily indexed table. Table definitions
and individual customizations are kept in a configuration
script and can be easily modified.

Software
WIRE is realized with robust, ubiquitous and inexpensive
software. The Swedish open-source product MySQL serves as its database
engine, whereas for internet querying the free web server
Xitami is used. The script
modules that combine both are written in Perl, a widely-used free
programming language. In the future it will be possible to
install the system and to create individual databases
through user-friendly routines. A special document template
for Microsoft Word helps to
input the data and to manage the database, so that
researchers won't have to leave their familiar working
environment. The formatting is converted into HTML code and
inserted into the database along with the unencoded text
which is used for searching. Except for the Word interface WIRE runs on other
operating systems, too.

5. The Users
WIRE is being designed for art historians, but it may be useful
in other disciplines of the humanities, too. It could serve
students and teachers alike, whether they work together in
research projects or on their own in seminars. Students could
present their papers using WIRE and at the same time preserve
and maintain the material for future use. Currently it is
already in use at Trier University as a platform for half a
dozen projects, some of which are dependent on international
collaboration.

Publications

Martin
Raspe

WIRE - ein Instrument zur Materialsammlung in
den Bildwissenschaften

EVA 2001 Berlin (Electronic Imaging &
the Visual Arts), Konferenzband

[forthcoming]

Title C: Towards the User: The Digital Edition of the Deutsche Wörterbuch by Jacob and Wilhelm Grimm
Ruth Christmann, M.A.
Thomas Schares, M.A.

1. Starting position, targets:
The Deutsche Wörterbuch (DWB) by
Jacob and Wilhelm Grimm is the most important dictionary to
the German language. Begun in 1838 and completed in 1971
with the publication of an index volume to the numerous
sources quoted within the DWB, it is a chief stock for
scholarly study of the German language and comparable to the
importance of the OED for the English-speaking world.
In November 1998, a team of lexicographers and computer
scientists started to develop a digitized version of the
DWB, taking into account the needs of academic researchers
wanting to cope with the huge amount of data. This is, and
always has been, a task not too easily performed, as the DWB
is used by students of German, historians, lexicographers,
and philologists of all disciplines.
The DWB consists of 32 volumes and one index volume. It fills
altogether 33,872 pages in folio-format, contains ca.
250.000 main entries, the number of printed characters
amounts to 300 million. (Compare the 2nd Edition of the OED:
21,730 pages, 231,000 main entries, 350 million printed
characters). The printed dictionary has been made
machine-readable by a Chinese company and has been completed
in October 2000. After having received the first files from
China we started to insert SGML-markup compliant to the TEI
guidelines into the dictionary. First we decided on marking
up two volumes that were published in the 1950s, as the last
volumes published have a fairly uniform structure.
Afterwards the procedures developed for these volumes were
applied to the other volumes successively, starting from the
first volume which had been published in 1854.
As a dictionary retrodigitization project, we do not have to
cope with different classes of information or media. What we
need to do is digitize dictionary entries, which exist
already in a final state and - what is by no means trivial
-, give users a firm grip on what they are looking for
within the dictionary without restraining them by supplying
unsuitable means for information retrieval.

2. Necessity for retrodigitization: Beyond the scope of
the printed dictionary
From the very beginning of the project, it was our aim to
anticipate the needs of the average user of the DWB. As the
DWB has complicated and heterogeneous structures as a result
of nearly one and a half centuries of philological research
and lexicography which determine the dictionary's contents
and structure, the access to the printed version of the DWB
is quite complicated: Not only does the user have to consult
more than one volume in most cases, as the DWB is full of
cross-references to other parts of the dictionary or to
certain columns within the same entry which may be found
dozens of pages apart from each other. The reader is also
confronted with the difficulty to find exactly those
paragraphs within an entry he is interested in, e.g. the
exact meaning he is looking for: Due to the long time it
took to complete the DWB, the entries are very
heterogeneously structured, and hierarchical elements vary
according to different underlying entry structures and often
serve different purposes.
Therefore, a digitized version of the DWB has to take into
account these problems and in the first place has to
facilitate a convenient access to the dictionary entries.
Second, the graphical user interface (GUI) has to support an
effective orientation within and a sophisticated navigation
through the dictionary, it should make it easy to follow up
cross-references within entries as well as within the
dictionary as a whole.

3. User-oriented approach to data via the DWB GUI
The DWB GUI is designed to make use of the special riches of
the dictionary and to provide a comfortable access to its
contents. It offers various possibilities for simple
headword search and provides a wide range of information to
chosen entries. It gives for example the exact reference
where in the printed dictionary is the entry located,
furthermore who is the author of the entry (more than 150
lexicographers participated in the making of the DWB) and
when did it appear in print (an important information for
evaluating the entry's contents with regard to historical
circumstances) etc. The - often cryptic - bibliographical
information within the dictionary referring to the sources
of the quotations is being made comprehensible by
interlinking it with the dictionary's index volume that
shows the sigla with their full bibliographical information.
A special feature of the DWB is the great number of very long
entries: the DWB has more than 50 entries consisting of more
than 50 columns and more than 200 entries consisting of 5 or
more columns, about 40 percent of the dictionary contents
consist of entries of five or more columns length. For ease
of reference, the GUI is provided with a window section
which visualizes the hierarchical structure of long entries,
especially in the sense section. For this purpose, the
various numbers and characters indicating sections and
sub-sections - there are up to nine sub-sections - are
ordered according to a structure resembling a family tree.
This overview is based on the far-reaching markup that takes
into account and describes the various functions of numbers
and characters representing the structural elements of an
entry.
This special feature enables users to enhance his search
strategies considerably by offering them this window with an
overview of the contents of an entry. From this table of
contents, users may look up the various 'chapters' of an
entry. A more detailed description of this special feature
shows the goal of the design of the DWB GUI: It will offer
an easy access to the dictionary contents by taking into
account and encoding all the essential features of the
dictionary.

4. The DWB retrieval mask
The electronic DWB will also be provided with a powerful
search retrieval tool. In addition to the basic features of
fulltext search and Boolean search, there is the possibility
of complex queries. These are carried out by making use of
the structural encoding of the dictionary's data. All key
elements and sections of an entry have been encoded
according to their structural positions in order to retrieve
as much information as possible.
At present, searches in the electronic DWB are still limited
to headwords, word class (part of speech), languages
especially in the etymology section, quotations from poetic
sources and bibliographical information on these.
Nevertheless, even now a user interested in word formation
and the author Goethe may start
a complex query to find all entries to adverbial derivations
with the suffix -lich which
contain at least one quotation by Goethe. The user will
combine the search for "G*the" (according to the variant
spellings "oe" or "ö") entered in the field for author/work
which is linked to the index volume, "*lich" in the field
for the headword, and "adv" in the field for word class. The
result will list all the relevant entries within the
dictionary. This may serve as an example for complex search
strategies which will be made possible in the electronic
version of the DWB.

5. The future of digitization and research
By now, volumes 1 to 9 and volumes 27 to 32 have been encoded
according to SGML/TEI and since Jan. 2002 been made
available on the Internet.
At the time of the conference, the complete DWB will be fully
encoded and accessible for searches described above.
Furthermore, PDF files have been designed to represent the
printed dictionary, these will also be accessible. We will
give some details as to the problems that had to be solved
when aiming at an exact representation of the character-sets
(Greek, Hebrew and others) on different platforms.
Questions to be discussed when presenting the DWB may focus
on future necessities of encoding the DWB and of data
encoding in general, especially in connection with the
question of user needs. This may also include a comparison
of the DWB to at least one major digitized dictionary on
historical principles, the Oxford English Dictionary (OED)
which is comparable in size and structure.

Literature

Thomas
Burch

Ruth
Christmann

Vera
Hildenbrandt

Thomas
Schares

Ein “Hausbuch” für alle? Das Deutsche Wörterbuch der Brüder
Grimm auf CD-ROM und im Internet

Jahrbuch für Computerphilologie

2

11-34
2000

Thomas
Burch

Kurt
Gärtner

Thomas
Schares

Das digitale Deutsche
Wörterbuch der Brüder Grimm

Mitteilungen des Deutschen
Germanistenverbandes

Forthcoming in 2002

Ruth
Christmann

Vera
Hildenbrant

Thomas
Schares

Ein “heiligthum der sprache” digitalisiert:
Das Deutsche Wörterbuch von
Jacob und Wilhelm Grimm auf CD-ROM und im
Internet

Nicolás
Castrillo Benito
et al
Tagungsband der ITUG-Jahrestagung 1999 in
Burgos: TUSTEP educa

Burgos

2002

Ruth
Christmann

Books into Bytes: Jacob and Wilhelm Grimm's
Deutsches Wörterbuch on
CD-ROM and on the Internet

Literary & Linguistic Computing

16
2
121-133
2001

Vera
Hildenbrandt

Thomas
Schares

Das Grimmsche Wörterbuch geht ins 21.
Jahrhundert: Präsentation eines Prototyps des digitalen
Deutschen Wörterbuchs von
Jacob und Wilhelm Grimm

().

Ruth
Kersting

Andrea
Rapp

Mein schönes Fräulein ... Bedeutungs- und
Bezeichnungswandel in Wortfeldern anhand des Grimmschen
Wörterbuchs

Praxis Deutsch

165

54-59
2001

Homepage: