The Meta Dictionary

Paper given at ALLC/ACH 2002, Tübingen, Germany

Christian-Emil Ore, Lars Jørgen Tvedt, Tone Bjørnstad
The unit for digital documentation,
Faculty of arts, University of Oslo 1123 Blindern, NO-0317, Oslo

In this paper we will present a lexicographic database tool called The Meta Dictionary. It is designed for systematising word and phrase information from field datasets for weakly normalised languages and dialects. The Meta Dictionary has been developed to function as a pivot in the revitalisation of a large traditional national dictionary project in Norway. In addition to completing the traditional dictionary manuscript, this revitalisation project has the objective of creating a modern lexical database in a way similar to what is now being done for The Oxford English Dictionary [10] or the The Cobuild Word Bank [4]. The Norwegian dictionary project was started in 1930, and the editors are now at the end of the letter H. According to the new plan this will be finished in 2014. The Meta Dictionary system is meant to be a general tool for working with lexicographic raw material and is not limited to the Norwegian dictionary project. The plan is to use it in a project on the minority languages in Zimbabwe in (2002-2205). However, it can also be adapted to implement semantical database based on different ontologies (semantic models) like the one used in WordNet [12] or in The Simple Project [11 ]. The Meta Dictionary and the other lexicographical tools described in this paper are developed by The Museum Project (MusPro). The MusPro was established in the spring of 1998 as a national collaborative project involving all four Norwegian universities. The objective is to develop common database systems for the meta data and the management of collections, accessible for all Norwegian University museums. The project organisation has taken over the responsibility for maintaining and developing the database systems for the department of lexicography (old Norse and modern Norwegian) and place name studies, and is a direct continuation of the systems development group in The Documentation Project a major digitalisation project in the 1990-ies.

Lexicographical background - the heterogeneity of a language

In Norway several linguistically oriented research groups participated in the development of electronic corpora from the beginning of the 1970's, (LOB, ICAME corpus projects, see [6]), but for unknown reasons the development of corpora consisting of Norwegian texts did not start before the end of the 1990's. As a parallel - there is still no complete and comprehensive national dictionary for Norwegian. Both cases have connections to the fluctuating language situation in Norway and the two closely related written Norweigan languages. On a world basis two written languages in a country is neither unusual nor impressive. What makes the Norwegian case unusual is the fact that two languages are completely comprehensible for all Norwegian speaking persons, that there are no clear- cut boundaries between the two, a rich flora of accepted alternative spellings and inflection schemas and finally frequent spelling reforms (although the period between these is increased to 4 years). The two written standards derived from written Danish (the only official written language in Norway from the sixteenth century until 1870) and on a synthesis of the dialects in the western parts of Norway. The latter was created by the linguist Ivar Aasen (see Norwegian Dictionary, 1872 [7]). During the last 130 years the standards have gradually become much closer. However, in non-official written communication many archaic forms are still in use. In private written communication people often mix the two standards. The language situation introduces many problems for the language technology industry and makes good solutions complex and expensive to develop. However, the relatively weak normalisation and the focus on dialects have a democratic effect, e.g. it is possible to use a written language pretty close to one's own spoken language. In most European countries normally thought of as a homogeneous language areas with a single official written language, there are in fact many distinct dialects or even related languages. The German language area (Germany, Austria and parts of Switzerland) is one example. Thus it can be argued that the language situation we find in Norway is the norm. However, the Norwegian language policy is more complex than may be expected. In recent years initiatives have been taken to at least complete the national dictionaries in Norway. It now looks as if there will be two printed dictionaries in respectively 7 and 12 volumes. The former will concentrate on the originally Danish based written language (used by most of the population) and will be based on an already existing dictionary in six volumes. The latter will describe the second written language and the Norwegian dialects although this is not planned as a dialect dictionary in the purest sense of the term. The work on this dictionary has already been running for 60 years. The editors are now working at the end of the letter H and at the current pace the dictionary will be finished in 2040. The Ministry of Culture has, pending the reorganisation of the team, and its completion by the year 2014, rather generously refunded the work. The new project is called Norwegian Dictionary 2014 (ND2014). From 2002 the editorial team will be increased from 6 to 28 full time editors. The main focus is on a method and tool for systematising source material and preparing the semantic skeletons of the entries. To explain the idea and the specific use of The Meta Dictionary we will briefly discuss our interdisciplinary database framework and the heterogeneous lexicographical source material.

Creating a multidisciplinary object oriented, database framework

To make database systems for the large number of disciplines is a challenge for a small group. An additional challenge is the requirement of providing for interdisciplinary searches. The number of databases and the aspect of inter-disciplinarity has forced us to try to make as generic database solutions as possible. The design and implementation of the common systems and interfaces are now completed in the first version. The new information system replaces older, mostly independent database applications. This is fortunate since it is easier to create new interconnecting systems instead of connecting old ones, although technologies like the Z39.50 standard have opened for the relatively easy interconnection of databases. The system group attempts to think generically along two axes:

Common interface tools and database functionality
Common database solutions for common object types like geographical data, bibliographical data, data concerning individuals and legal bodies, classification systems in cultural and natural history, dictionary entries and so on.

The databases are event oriented. The information is considered to be the result of observations and all major changes result from a well-defined event. The event -based object oriented model is also used in the tools now designed for the dictionary project ND2014. The system consists of three parts: 1) The databases containing the background material, that is, databases containing the old slip collections, dictionary manuscripts, bibliography etc., editors notes and the corpora of older literary text and modern Norwegian. 2) The dictionary editing system 3) the backbone of the system, The Meta Dictionary itself.

The Lexicographic source material

In the late eighteenth century, paper slip collections were introduced in dictionary making. A single word form with an example of actual usage was written on each paper slip. The slip collections were stored in alphabetic order in large filing cabinets. This was a major achievement and introduced scientific methods into lexicography. From the end of the 18th century most large-scale dictionary projects, e.g. The Oxford English Dictionary (OED, first edition completed in 1930), have been based on slip collections such as this one. At the outset of 1960 one started to use computers to create concordances from electronically stored texts and the use of slip collections was regarded increasingly as being old fashioned. Over the last 10 years the number of corpora has exploded: there are now text corpora for large number of languages and social groups. It is almost impossible to imagine that someone should start building a paper slip collection today. However, many of the large dictionary projects initiated in the first half of the 20th century and even earlier are not completed and still rely on their slip collections (The Afrikaans Dictionary, WAT, The Dictionary of the Swedish Academy, Dutch national dictionary, to mention a few). The two major Norwegian dictionaries were planned in the tradition of the large historically oriented, national dictionaries like OED and during the years slips were collected (numbering more than 6 millions today). In the 1990 most of the old collections were made electronically available either as indexed electronic facsimiles (3.2 millions) or replaced by collections of SGML tagged electronic texts. A series of older dictionaries and historical word lists from the end of the seventeenth century and onwards was also made electronically. This was done by the forerunner of The Museum Project, The Documentation Project, and has been described at ALLC/ACH 1995 and 1996 ([5][9]).

The editing system

A standard electronic editing system for a dictionary manuscript is based on a database in which the entries and parts of the entries are stored. Each entry can be stored in one or many tables, e.g. separate tables for the headword parts, the definitions and the citations, or the entire manuscript can be stored in a XML based free text system. A typical editing system usually supports cross-referencing, searching and the production of camera-ready page lay out. MusPro has made several such freestanding tools for a national language project in Zimbabwe (Shona and Ndbele)[2] and for one volumes dictionary in Norwegian. The latter are both free standing and connected to the common framework and the Meta Dictionary.

The Meta Dictionary

The source databases/datasets (described above) contains samples from different dialects and times. The most important lexicographical task is to systematise the material into groups, each describing a descriptive word. This process is called the lemma/headword selection. For strongly normalised written languages like Standard English this is normally not considered a difficult task. For field lexicographers working on not so normalised languages, this is well known as difficult task. The question "what is a word "and how to select the right candidate among a set of variants is not a trivial task. For most existing languages the standardisation has been mostly based on socio-political circumstances. The Meta Dictionary system has no features for (semi)automatic lema selection but is designed to assist the lexicographers in the process, to make it easier for others to check the lexicographers' decisions and open for more than one normalisation based on the same material.

In this process we have tried to exploit the interdisciplinary nature of our organisation
to make a tool for preparing the entries in the Norwegian Dictionary
to create a general tool for systematising information from weakly normalised languages and /or dialect material into a lexical database.

The Meta Dictionary is based on the general system for the users' electronic notebooks or folders used in all our database-applications. In general all our users can store parts of the result from a query for later use in their private notebooks in the system. The notebooks are multi-typed and can store information of all the object types in the databases. Traditionally, a working lexicographer usually sorts the available material first under each headword and then according to the meanings and use before formulating the definitions. In The Meta Dictionary an entry is a folder containing samples use taken from the source databases and corpus system connected to it. The user uses the drag and drop mechanism to move pieces of source information into the folder. Each entry/folder is labelled by the normalised word(s) and word class information and information about the actual orthographically standard used. The database used by the editorial team of the Collins Cobuild dictionary in the lemma selection of from the corpus is an early example of such a database. Several standards can be used simultaneously in The Meta Dictionary. This is useful in the Norwegian context. One possible application could for example be a Meta Dictionary for German dialects having a Low German and a High German normalisation. A more realistic application of The Meta Dictionary system is to use it as a descriptive tool for old written languages like Old Norse. It is planned to be used in systematising the data from the fieldwork on minority languages in Zimbabwe in the phase 3 of the The ALLEX Project (2002-2005) [2]. In the Norwegian dictionary (ND2014) project The Meta Dictionary system has been used during the last 2 years and more than 150 000 entries out of a total estimated to 450 000 have been created. The Meta Dictionary can be seen as an amalgamation of a lexical database, an electronic slip collection, a set of older dictionaries and a editing tool or database with a growing manuscript for ND2014. The backbone of The Meta Dictionary is the list of all the headwords from A to Å (the last letter in the Norwegian alphabet). The other information is linked to this list. The source materials in the entries/folders can be view in their original layout. That is, all the samples can be displayed "inline" as they are in their separate databases.

Connecting the editing of the dictionary and The Meta Dictionary

The basic Meta Dictionary is a tool for group the background information (samples) to (head)words. In a (monolingual) dictionary the defining parts of the entries usually are divided in a tree like structure reflecting the semantics of the words. Thus The Meta Dictionary used in the NO2014-project is linked to the editing system by a drag and drop based interface for sorting the information in The Meta Dictionary entry/folder into a tree-like hierarchy which in turn can be view by the editor in the editing system as the skeleton of an ordinary entry in ND2014. By these means it is possible to survey back and forth between the collected source material and the dictionary manuscript. These links will be kept after the editing of an entry is finished. Furthermore, the entry for a given word will automatically be placed in The Meta Dictionary entry/folder. The NO2014-entry will "live" together with its sources and in fact be just another information piece describing that word. In a corresponding way it is relatively easy to connect/extend The Meta Dictionary to WORDNET-like descriptions or any other ontology-based semantic structures. The users of The Meta Dictionary connected to ND2014 do not need to have a ND2014 perspective on the word information, but may freely chose their focus. The information given in ND2014 is simply one out of many seen in a time perspective. We believe that the The Meta Dictionary also is well suited to create a general tool for systematising information from weakly normalised languages and /or dialect material into a lexical database.


[1]Making an interdiciplinary system, Christian-Emil Ore ALLC/ACH 1993, Georgetown University, Georgetown, USA
[2]The African Languages Lexicography Project (The ALLEX project)
[4]The COBUILD dictionary and pointers to the project:
[5]"New life for old reports - The Archaeological Part of the National Documentation Project of Norway", Jon Holmen, University of Oslo at ALLC/ACH 1996, University of Bergen, Bergen, Norway
[6]HIT (Programme for Information Technology in the Humanities),
[7]Norsk Ordbok [Norwegian Dictionary], Aasen,I. Christiania, Norway, 1872
[8]Norsk Ordbok [Norwegian Dictionary], Volumes I-IV, 1950-2001
[9]"The Corpus and the Citation Archive - Peaceful Coexistence Between the Best and the Good?", Christian-Emil Ore, University of Oslo at ALLC/ACH 1995, University of Santa Barbara, Santa Barbara USA
[10]Oxford English Dictionary, Oxford 1928 and
[11]The SIMPLE project,