967Lexicography in gLobaL contexts
Advances in Synchronized XML-MediaWiki Dictionary
Development in the Context of Endangered Uralic Languages
Mika Hämäläinen, Jack Rueter
Department of Digital Humanities, University of Helsinki
E-mail: mika.hamalainen@helsinki., jack.rueter@helsinki.
We present our ongoing development of a synchronized XML-MediaWiki dictionary to solve the problem of
XML dictionaries in the context of small Uralic languages. XML is good at representing structured data, but it
does not fare well in a situation where multiple users are editing the dictionary simultaneously. Furthermore,
XML is overly complicated for non-technical users due to its strict syntax that has to be maintained valid at
all times. Our system solves these problems by making a synchronized editing of the same dictionary data
possible both in a MediaWiki environment and XML les in an easy fashion. In addition, we describe how
the dictionary knowledge in the MediaWiki-based dictionary can be enhanced by an additional Semantic Me-
diaWiki layer for more eective searches in the data. In addition, an API access to the lexical information in
the dictionary and morphological tools in the form of an open source Python library is presented.
Keywords: online dictionary, collaborative editing of XML, Semantic MediaWiki dictionary
In this paper, we present advances in the development of our open-source synchronized XML-Me-
diaWiki dictionary environment1 (Rueter & Hämäläinen 2017). The dictionary data consists of mul-
tiple XML dictionaries for small Uralic languages2 following the same XML structure. XML diction-
aries are used on the Giellatekno infrastructure (Trosterud, Moshagen & Pirinen 2013) for many dis-
tinct facets of linguistic research such as Intelligent Computer-Assisted Language Learning (ICALL)
(Antonsen et al. 2014), FST generation for morphological analyzers and spellcheckers.
XML is a great format for storing structural data, such as information usually stored in a dictionary.
It does, however, have some drawbacks, such as editing XML data in a collaborative fashion is a
challenging task. This is even more so in the case of non-technical people native in endangered Uralic
languages. In order to enable them to produce and correct dictionary resources, a simplied way to
edit XML data is needed.
We have thus developed a MediaWiki-based online dictionary system, the purpose of which is to
make it possible to edit structural dictionary data collaboratively with a simplied interface. The dic-
tionary works in such a way that we can get the edits instantly in our XML formalism, and edits made
directly in the XMLs are also updated to the MediaWiki.
1 Available on https://sanat.csc./
2 The languages currently supported in the dictionary are Skolt Sami, Ingrian, Meadow Mari, Votic, Olonets-Karelian, Erzya,
Moksha, Hill Mari, Udmurt, Tundra Nenets and Komi-Permyak
968 Proceedings of the XViii eUrALeX internAtionAL congress
2 Related Work
This section presents some of the previous research done in the context of online dictionaries. The
previous work ranges from theoretical takes on online dictionaries to actual online systems imple-
mented for the task. In a meta-analysis of studies on the usage of electronical dictionaries (Töpel
2014), several advantages in electronical dictionaries were identied. A positive impact on speed,
performance, ease of use, vocabulary retention and satisfaction were reported in a dictionary use
The XML structures of this project are compatible with and, where possible, identical to those used
by the dictionaries in the Giellatekno infrastructure, where local enhancement provides the availa-
bility of special glyphs for assistance in individual language input, links to corpora search in Giel-
latekno-hosted Korp, as well as grammatical links for the enlightenment of the lay user3. Whereas
the Giellatekno dictionaries provide for dictionary users without specic keyboards for the individ-
ual languages, we require our users to have keyboards of their own4. Pointer data in our XML and
MediaWiki interfaces allow us to open individual page links to the etymological database for Sami
languages (Álgu-tietokanta 2002).
Uralic language databases are the target of continuous development in Estonia. This can be observed
in the outline of Estonian and Uralic language archive materials in Tallinn and Tartu (Viikberg 2008),
and subsequent mention of work on Estonian-Mari and Estonian-Erzya dictionaries at EKI (Eesti
Keele Instituut [Estonian Language Institute]) in EELEX (Tender et al. 2017). Similar bilingual dic-
tionary development with audio resources are described for Võro-Estonian (Männamaa & Iva 2015).
The Dictionary of Old Norse Prose (ONP) (Johannsson & Battista 2017) has implemented multiple
search and presentation features. It strives towards an online tool with enhanced corpus search and
allows for presentation of manuscript and archive materials, as well as individuated download possi-
bilities. In our project, however, we retain a light structure with synchronic editing for XML and Me-
diaWiki. The XMLs contain a set of hand-selected example sentences from corpora to be displayed to
the user in the online dictionary. However, our system has not been linked to full corpora for example
The role of e-lexicography is growing. Not only is the detail required for the conversion from printed
dictionaries to digital format being examined, but investigations are also being made of the feasible
saturation of data presentation. E-lexicography allows for the introduction of new tools, and is seen
as an opportunity to provide direct data extraction from various data sources (Bothma, Gouws &
Prinsloo 2017). Our MediaWiki presentation involves three dimensions of linking. It includes links
to external datasets (etymology and audio), other languages in the internal dataset (denitions, ety-
mology), and dictionary internal links between articles (compound word constituents and derivation
stems). We also generate regular paradigmatic tables for viewing while retaining a view of lemma,
native denition, translation and morphologically important category information on the screen.
3 Giellatekno online morphologically savvy dictionaries with click-in-text readers and possible Korp links are available at: http://
sanit.oahpa.no/ (North Sami), http://baakoeh.oahpa.no/ (South Sami), http://saanih.oahpa.no/ (Inari Sami), http://saan.oahpa.no/
(Skolt Sami), http://sanat.oahpa.no/ (Northern Balto-Finnic languages), http://sonad.oahpa.no/ (Southern Balto-Finnic languages),
http://valks.oahpa.no/ (Mordvin languages), http://muter.oahpa.no/ (Mari languages), http://kyv.oahpa.no/ (Permic languages),
and http://vada.oahpa.no/ (Nenets).
4 The necessary keyboards for most Uralic languages are produced for Windows, Mac and Android and available at http://divvun.
no/ for Saamic languages, and analogical keyboards for other languages can be generated directly in the Giellatekno infrastructure.
LexicoGraPHy iN GLobaL coNtextS
3 The XML Dictionaries
The XML dictionaries draw upon the goal of minimizing data redundancy in di erent branches of
an extended infrastructure at Giellatekno (Trosterud, Moshagen & Pirinen 2013). Original paral-
lel sources existing for online morphologically savvy translation dictionaries, on the one hand, and
minimal sized ICALL dictionaries, on the other, have been integrated with lemma:stem pair data
utilized in transducer production. Subsequently, other research data has been incorporated into the
XML structure as well, such as audio pointers, and etymological as well as derivational informa-
tion partially inherited from previous language projects. Thus, while the dictionaries can be used
through XSL transformation to provide code output (lexc) for the construction of transducers used
in nite-state morphological analyzers and spell checkers, they also serve as extensive databases for
other research projects. The distinction between source and target languages is maintained utilizing
ISO 639-3 three-letter codes, which can be attested in the XML root element as well as the translation
group <tg/>, example group <xg/>, etymon and cognate elements.
The translation dictionaries were originally set up as source-to-target, bi-lingual dictionaries. In word
entries with broader semantic coverage, granularity has been introduced. This allows for multiple
translations in the instance of semantically close de nitions <t/>, and separate meaning groups <mg/>
for distinct senses of a word. Contextual usage is demonstrated in Figure 1, which shows translation
groups within the semantically appropriate meaning group. The Giellatekno dictionaries based on
this XML structure are available and undergoing continuous development within the Giellatekno
Figure 1: XML entry
Figure 1: XML entry
Optional enhancement of the underlying lemma (e/lg/l), stem (e/lg/stg/st) and in ection (e/lg/stg/
st@Contlex) dictionaries can be observed in the etymon and audio pointers, as well as the deriva-
tion (e/lg/compg), translation (e/mg/tg) and example (e/mg/xg) groups. While the lemma, stem and
continuation lexica data serve as vital information in tranducer development, etymology, audio and
compounding pointers provide for navigation between and within dictionaries. The etymon pointer
ProceediNGS oF tHe xviii euraLex iNterNatioNaL coNGreSS
is used to access an online Sami language etymology dictionary, whereas an optional cognate sibling
allows for pointing between languages in the MediaWiki infrastructure. Likewise, the inc-audio/
audio pointers allow for accessing recordings in the Max Planck Institute archives at Nijmegen, and
compounding group pointers o er access for navigation within the source language at the lemma and
su x levels.
Homonymy is addressed on a part-of-speech basis, with words bearing mutual etymological and in-
ectional data subordinated to single entries but feasibly di erent senses.
4 The Synchronized Dictionary System
The synchronized dictionary system we are proposing is meant to solve the problem of collaborative
editing of XML dictionaries. Having multiple editors modifying the contents of pure XML diction-
aries simultaneously is not an easy task to accomplish. It gets even more di cult if the editors have
only a very limited technical background and from little to no understanding of the XML syntax.
Large-scale tasks such as crowdsourcing of dictionary editing become next to impossible with plain
Another XML speci c limitation our system is made to solve is breaking the tree structure of XML.
Our dictionary system can build links in between di erent lexical entries even across multiple dic-
tionaries to provide a more graph like structure of the dictionary data. This also makes it possible to
conduct more complex queries to search for information stored in the dictionary system.
What makes our system synchronized is that we do not want to move entirely away from the XML
standard, but rather build a system in which the same dictionary information can be edited in an easier
crowdsourced fashion in a MediaWiki environment and also directly in the XMLs, so that edits at
either end of the system will be made instantly available to all viewers of the dictionary system.
Figure 1: System architecture
Figure 2: System architecture
The core of the dictionary system is the Django-based Synchronization DB seen in the middle of Fig-
ure 2. This database application provides APIs to the Git integration for making changes to the XML
les and also APIs to the MediaWiki integration to communicate with the Wiki environment. The
di erent parts of the system are discussed in more detail in the following subsections.
The non-functional requirements for the system are reliability and scalability. The dictionary system
should not only serve researchers but also any language user outside of academia, which is a reason
971Lexicography in gLobaL contexts
why reliability of the platform is needed to guarantee a decent uptime. One of the design principles
is that we should be able to include new dictionaries in the system, which means that it should scale
well. The system should also be built fully on open source technologies in order to ensure its compat-
ibility and maintainability in the future. Because the underlying MediaWiki platform and the XMLs
will be used for other purposes than those required by our system, we also need to follow the idea of
separation of concerns in order to ll a criterion of integratability.
4.1 Synchronization Database
The role of the synchronization database is to keep the most up-to-date version of the data in all
situations. This makes it possible to isolate the synchronization feature from the XMLs and the Me-
diaWiki, making it possible to introduce new sources and views to the data in the future. These might
be XMLs following a dierent structure or an entirely new system for collaborative editing. By em-
bracing the notion of separation of concerns, we do not want to build the synchronization database
to follow the structure of MediaWiki syntax or XML syntax, but rather we want it to have its own
When constructing the system, we want to keep the option open for introducing new data sources,
in XML or in another format. This means that the contents of the data are not predictable, and thus
dening an SQL database would make incorporating new kinds of data dicult. Storing data in plain
XML format is not a viable option either, as using an XML database has a huge negative impact on
the performance of the system (Nicola & John 2003).
We thus propose using MongoDB as a solution for storing the data in an eective fashion. MongoDB
is a so-called NO-SQL database which does not require a predened structure for the database. In
performance terms, it can run faster than a traditional SQL database (Boicea, Radulescu & Agapin
2012), making it a good option for our purposes.
The database application is a Django-based web application. Communication from the Git side and
MediaWiki side with the database is thus done by using HTTP requests to the web application API.
The process of bringing XML les to the system is done over Git.
When XMLs are edited or brought for the rst time to the system, the changes made in them are com-
mitted to a Git repository. If the XML dictionaries already exist in the system, the editor has to run a
special command line script that will create a new branch and dump the data from the synchronization
database into XML format in that branch. This leaves the conict resolution to the editor of the XML
les. He can compare his current working branch with the latest data in the synchronization database,
resolve the possible conicts and merge the branches to the master branch. When the repository is
pushed, the synchronization database pulls the changes and updates its internal database after which
it starts updating the MediaWiki side.
The XMLs are read into the internal JSON format of the system by language and/or XML structure
specic modules. When the XMLs are requested from the system, a format and language specic
Django template is used to produce the XML structure. This conversion process of the data is ex-
plained in more detail in the next section.
4.2 Support for Multiple Languages
The system is built in a modular way to facilitate the inclusion of new languages or data sources. At
the moment, all of the languages in the system follow the Giellatekno XML syntax, which means
that the same modules are reused just with a dierent language ag. The system needs two language
modules, one to handle the XML to JSON conversion for MongoDB and another to handle the JSON
972 Proceedings of the XViii eUrALeX internAtionAL congress
to MediaWiki syntax conversion. We are dealing with languages whose orthographies contain special
characters. This means, for multiple language support, that we have made sure the data is handled in
UTF-8 format in all parts of the process.
Since the synchronization database itself is unaware of the contents of the data, how the XML gets
transformed into JSON can be decided for each language module separately to better suit the needs
of each dictionary type. The module currently developed for the Giellatekno XML does quite a direct
transform of the XML data into JSON format. We do, however, handle homonyms dierently in the
JSON. In the Giellatekno format, homonyms are completely separate entries in the XML with a hid
attribute to indicate the ID of the homonym. In the JSON format, it will be noted, we include all hom-
onyms in a list under the main entry, which is identied by the lemma. The reason for this is simple:
on the MediaWiki side all the homonyms are listed inside the same article which is identied by the
lemma. Having all the dierent homonyms in the same entry in the synchronization database makes
producing a MediaWiki page much simpler.
The other part of the language module is a script that can be run both in the synchronization system
side and in the MediaWiki side to do a conversion between JSON and MediaWiki syntax. The im-
portant part is that the MediaWiki syntax is only used for the visualization of the dictionary data.
For editing the dictionary entries in the MediaWiki side, a dump of the JSON data is included in the
article in a hidden div element.
4.3 MediaWiki Integration
The MediaWiki integration is an extension which is isolated to work with a predened set of name-
spaces. Our system creates a new MediaWiki namespace for each language. In practice, this means
that each entry is prexed by a three letter ISO language code, for example the Skolt Sami word sokk
is stored inside of the MediaWiki article named Sms:sokk. The reason why it is important to limit the
functionality with namespaces is not only that the namespace tells which language module should be
used, but also that our dictionary system is a part of a shared MediaWiki dictionary of the Language
Bank of Finland with multiple dierent data providers. This additional namespace restriction makes
sure that our solution does not interfere with the MediaWiki entries other projects are building.
The MediaWiki extension of our system, in addition to communicating the changes to the synchro-
nization database, provides the functionality for two MediaWiki article views: visualization and ed-
iting. A language specic module is used to construct a viewable version of a dictionary entry, or an
article in the MediaWiki terminology. As described before, this viewable version stores the JSON
structure as a hidden element for editing purposes.
The editing part of the MediaWiki extension solves the problem of XMLs requiring additional techni-
cal knowledge to be edited. The edit view of a MediaWiki article hides the MediaWiki syntax editor
that would be shown by a MediaWiki based system by default. Instead, the editor constructs a form
based on the language module and the hidden JSON element as seen in Figure 3. When we force users
to edit the data through a form, we can make sure that the data is in a valid, parseable format. There is
thus no possibility for the user to accidentally break the syntax of the data structure by, for example,
forgetting a closing tag. Additionally, using a form for editing makes it possible for us to do form
validation before saving the data in the system. At the moment, the validation means removing empty
entries, such as a language entry without any translations.
Saving the edit form makes the system update the hidden JSON element and reconstruct the edit view
based on the new JSON data using the exact same functionality as when a synchronization database
pushes a JSON entry to the MediaWiki side. New changes are then immediately communicated to the
synchronization database through the MediaWiki extension.
LexicoGraPHy iN GLobaL coNtextS
Figure 1: Form in MediaWiki
Figure 3: Form in MediaWiki
Since the MediaWiki stores each dictionary entry as a separate article, and the synchronization data-
base does a similar separation, collaborative editing is made possible. Changes can be communicated
between the two systems per entry basis without the need to parse an entire collection of lemmas, as
in the case of XML. This structural separation of entries means that if di erent dictionary entries are
edited simultaneously, there will not be any con icts, but multiple edits can be synchronized in real
time. The only case of simultaneous editing that is not supported is when the same MediaWiki article
is edited at the same time by multiple users.
In addition to editing and visualizing the data, the MediaWiki integration has a search functionality
for accessing the dictionaries. This is needed because the MediaWiki environment contains so many
di erent dictionaries and word lists that using the default search box provided by MediaWiki makes
it next to impossible to nd the words in the system for an average user who is not familiar with the
namespacing used in the system.
Figure 1: Easy search interface
Figure 4: Easy search interface
ProceediNGS oF tHe xviii euraLex iNterNatioNaL coNGreSS
The simpli ed search interface is depicted in Figure 4. It provides the functionality of picking the
dictionary in which the words are searched, such as the Skolt Sami dictionary. Due to the highly in-
ectional nature of Uralic languages, a language learner might come across with a non-lemmatized
form of a word. For this reason, our search interface incorporates morphological analyzers to lemma-
tize the user input word form. As seen in Figure 4, the search term used was soʹǩǩe, and the system
found that it is an in ectional form of suukkâd, sookkâd and sokk. The inclusion of this feature is also
motivated by previous research (Bergenholtz & Johnsen 2005) pointing out that online dictionary us-
ers use non-lemmatized word forms (the passive and imperative forms of a verb in their study) when
consulting a dictionary.
It is also possible to use the same search to nd words in the translations. This means that by input-
ting the English word row, the system will nd the Skolt Sami entry suukkâd. The simpli ed search
interface also provides a link to the full MediaWiki entry.
4.4 Semantic MediaWiki
Semantic MediaWiki (Völkel et al. 2006) is an extension that has been used in the past in the Lan-
guage Bank of Finland MediaWiki environment with good experiences in the context of online dic-
tionaries (Laxström & Kanner 2015). The extension makes it possible to link MediaWiki articles
together based on shared semantic characteristics. The aim of the extension is to make semantic
knowledge in a MediaWiki environment machine readable.
We use Semantic MediaWiki to gain access to a more graph-like representation of the dictionary data.
We use it to enhance the MediaWiki entries with property tags in an automated fashion. The property
tags are added or updated to the MediaWiki articles automatically always when new edits are made.
Figure 1: Semantic MediaWiki search
Figure 5: Semantic MediaWiki search
The property tags such as tr_eng or Contlex make it possible to query the dictionary information more
e ectively through the Semantic MediaWiki query interface. In Figure 5, we see how we can get a
list of all Skolt Sami (Lang::Sms) words that do not have an English translation (tr_eng::no) and are
975Lexicography in gLobaL contexts
verbs (POS::V). We can also specify the property values we want to be visualized in the search results
such as continuation lexicon (Contlex) and the assonance rhyme structure of each word (Assonance).
These queries can be made within one dictionary or across multiple dictionaries stored in the system
by altering the Lang:: query parameter.
Furthermore, the extension allows us to access other entries of the same dictionary or entries of com-
pletely dierent dictionaries in the same system. This is achieved with the pages that link here func-
tionality. This means that we can see, for each entry in the dictionary, if there is another entry possibly
even in a dierent dictionary making a reference to a specic entry. Currently, these references might
be translations, derivations or etymologies. In other words, just by having an etymological relation
dened in the Skolt Sami dictionary, we can see the reference in the Erzya dictionary, for instance.
4.5 The API
As the dictionary uses morphological tools for dierent tasks, such as producing inection paradigms
when viewing an article in MediaWiki or lemmatizing input words in the simplied search view, the
dictionary system has in built functionality that can be of a general interest when doing NLP for Ural-
ic languages. This is the reason why we have decided to serve the morphological tools over an API
that is currently usable through a Python library called Uralic NLP5 (Hämäläinen 2018).
The underlying functionality relies on nite-state transducers based on the HFST tool (Lindén et al.
2013). These are openly available in the Giellatekno infrastructure (Trosterud; Moshagen & Pirinen
2013) in a source code format. Our API provides easy access to precompiled versions of the FSTs
for morphological analysis, generation and lemmatization. In addition to the FSTs, the API makes it
possible to get full JSON entries for words in the dictionary.
Apart from our own extended API, the standard MediaWiki API and Semantic MediaWiki API are
available for the users. These provide a standardized access to the data stored in the MediaWiki side
of the system, such as using the Semantic MediaWiki query language.
5 Lexicographical Dierence of the XMLs and MediaWiki
Each dictionary is tailored to a dierent audience or user group. Whereas the XML dictionaries have
been set up to act as virtually stand-alone databases that can be used for deriving any variety of output
sets, the MediaWiki dictionaries have been set up to provide a less cluttered experience. In fact, the
visible code in the MediaWiki presentation is less than what can be found in the XMLs. This design
decision was taken to better support the end user goals when using the dictionary. A typical diction-
ary user is more likely to be interested in denitions and translations than metadata or FST specic
information needed to produce the morphological analyzers. Visualizing too much information that
is irrelevant for the user goals makes it harder for the user to nd the relevant pieces of information.
This would cause higher cognitive load which would take up more working memory (Paas, Renkl
& Sweller 2003), which is the very thing we want to avoid with our design choice. Previously it has
also been reported that extremely extensive entries cause diculties in using the dictionary (Selva &
The MediaWiki dictionaries utilize three dierent types of links. Etymon and audio links provide
access to sites of external institutions, such as the Sami-language etymological database Álgu at the
Institute for the Languages of Finland in Helsinki, and the Max Planck Institute audio archives in
5 Instructions and installation on https://github.com/mikahama/uralicNLP
976 Proceedings of the XViii eUrALeX internAtionAL congress
Nijmegen. Cognate links facilitate navigation between languages in the namespace of our project on
the CSC/Language Bank server, while compounding and derivation links enhance the navigation ex-
perience between compound words and their constituents in the same manner as derived words point
to their derivational stems and morphemes. This interlinking provides a new alignment of semantic
and morphological data not immediately accessible from the XML databases.
Not all homography is dealt with by means of Roman numeral identication. In fact, the development
of XML dictionaries has led to the separation of homographs according to part-of-speech designation.
When the MediaWiki dictionaries return all homographs to adjacent micro-entries within the mac-
ro-entries, micro-entries with the same part-of-speech designation are distinguished, as in the XML
dictionaries, according to homograph enumeration, while other instances of homography are simply
addressed with the help of part-of-speech marking.
Semantic tag values with synset distinctions are used in some language development at Giellatekno.
In anticipation of shared meaning groups in source-to-multi-target-language dictionaries, this initial
semantic tagging has been introduced in the XML dictionaries, where they reect the same semantic
tagging used in Constraint-Grammar disambiguation applied in the Giellatekno and Apertium infra-
structures, and the ICALL infrastructure at Giellatekno. Initial outlines have also been drafted for
editing semantic links that will enhance searches for various degrees of synonymy.
6 Discussion and Future Work
Our system is under continuous development, but it has reached a functional state. At the moment,
we have several authors editing the Skolt Sami and Erzya dictionaries in the MediaWiki environment,
while part of the dictionary editing is still ongoing in the XMLs. In this case of a handful of editors,
the system has proved functional. The biggest limitation in the system, however, is the Semantic
MediaWiki extension. Enabling the extension has a huge impact on the speed of the system when up-
dating the entries in the MediaWiki side. We are currently nding ways to overcome this limitation.
The development has focused mainly on the technical side of the environment. Since the system is
meant to be used by people with no linguistic or technical background, more research is needed in
terms of usability and user experience of the system. This is especially needed and, in general, under-
studied in the context of editing the dictionary entries.
Giellatekno XMLs have the problem that they are not standardized by any means. This could be
solved by remodeling the XML structure in a standardized TEI format. Since our system is built with
multiple XML formalisms in mind, introducing a new TEI based format should not be too big of an
issue. In fact, by writing a new template we can already start producing a TEI formatted version of
the XML data.
The non-functional requirements of the system, reliability and scalability were solved by building the
system on industry-scale open source technologies. These are MediaWiki, Django and MongoDB.
Although these individual components are known to work reliably and scale well, there is a future
problem of maintainability. This rises from the concern of the compatibility of our system with the
future versions of MediaWiki and Django. Even during the two years we have been developing the
system, a critical part of the MediaWiki API has already changed once. This required updates to our
code in order to make our system work with the latest version of MediaWiki. This maintainability
issue is solved by releasing the entire system as open source.
Currently, other users of the shared MediaWiki platform maintained by the Language Bank of Fin-
land are showing interest in our system. Not only because it provides an already implemented way of
977Lexicography in gLobaL contexts
pushing dictionary data from another format to the MediaWiki system, but also because our system
makes it possible to transfer the data edited in MediaWiki back to the original format.
In this paper, we have described our online dictionary system6 with the aim of making XML based
dictionaries editable by multiple users. We have described the advantages and limitations of Sematic
MediaWiki in enhancing access to the dictionary data. Furthermore, the advantages of MediaWiki
have been described. Our system is currently in use and has been proved to solve the problems we
were set to solve with a small number of editors.
The dictionary system was originally developed for Skolt Sami, but we have successfully expanded
it to cover 10 additional languages with minimal modications. This has been possible due to the
modular nature and ideology of separation of concerns embraced in the design process.
In addition to solving a dictionary editing problem, our eorts have made the XML formatted dic-
tionaries available to a wider audience in an open MediaWiki format. The availability of these lexical
resources online has a direct impact on the speakers and learners of these minority languages. The
data has also been made available for research and technical purposes through the API of the system.
Álgu-tietokanta. (2002). Retrieved March 2018, from Kotimaisten kielten keskus: http://kaino.kotus./algu/
Antonsen, L., Johnson, R., Trosterud, T. & Uibo, H. (2014). Generating Modular Grammar Exercises with Fi-
nite-State Transducers. Proceedings of the second workshop on NLP for computer-assisted language learning
at NODALIDA 2013, (pp. 27-38).
Bergenholtz, H. & Johnsen, M. (2005). Log Files as a Tool for Improving Internet Dictionaries. HERMES-Journal
of Language and Communication in Business, 34, 117-141.
Boicea, A., Radulescu, F. & Agapin, L. I. (2012). MongoDB vs Oracle - database comparison. Third International
Conference on Emerging Intelligent Data and Web Technologies (pp. 330-335). IEEE.
Bothma, T. J., Gouws, R. H. & Prinsloo, D. J. (2017). The Role of E-lexicography in the Conrmation of Lex-
icography as an Independent and Multidisciplinary Field. Proceedings of the XVII EURALEX International
Congress, (pp. 109-116).
Hämäläinen, M. (2018, January). UralicNLP (Version v1.0). Zenodo. http://doi.org/10.5281/zenodo.1143638.
Johannsson, E. T. & Battista, S. (2017). Editting and presenting complex source material in an online dictionary: the
Case of ONP. Proceedings of the XVII EURALEX International Congress, 117-128.
Laxström, N. & Kanner, A. (2015). Multilingual Semantic MediaWiki for Finno-Ugric dictionaries. Septentrio
Conference Series, 2, pp. 75-86.
Lindén, K., Axelson, E., Drobac, S., Hardwick, S., Kuokkala, J., Niemi, J., Pirinen, T. & Silfverberg, M. (2013).
HFST — A System for Creating NLP Tool. International Workshop on Systems and Frameworks for Compu-
tational Morphology, (pp. 53-71).
Männamaa, K. & Iva, S. (2015). Võro-eesti-võro võrgosõnaraamat: synaq.org. In M. Velsker, & T. Iva, Tartu
Ülikoolo Lõuna-Eesti keele- ja kultuuriuuringute keskuse aastraamat (p. 147−150). Tartu: Tartu Ülikooli
Nicola, M. & John, J. (2003). XML Parsing: A Threat to Database Performance. Proceedings of the twelfth interna-
tional conference on Information and knowledge management (pp. 175-178). ACM.
Paas, F., Renkl, A. & Sweller, J. (2003). Cognitive Load Theory and Instructional Design: Recent Developments.
Educational Psychologist, 38(1), 1-4.
6 The system has been released as open source in https://bitbucket.org/mikahama/saame/
978 Proceedings of the XViii eUrALeX internAtionAL congress
Rueter, J. & Hämäläinen, M. (2017). Synchronized Mediawiki Based Analyzer Dictionary Development. The Third
International Workshop on Computational Linguistics for Uralic Languages, (pp. 1-7).
Selva, T. & Verlinde, S. (2002). L’utilisation d’un dictionnaire electronique: une etude de cas. Proceedings of the
tenth EURALEX International Congress, (pp. 773-781).
Töpel, A. (2014). Review of research into the Use of Electronic Dictionaries. In C. Müller-Spitzer, Using online
dictionaries (pp. 13-54). Berlin - New York: De Gruyter.
Tender, T., Kallas, J., Laansalu, T., Nurk, T., Mihkla, M., Päll, P., Langemets, M., Soon, T. & Oro, K. (2017). Eesti
Keele Instituudi osakondade aruanded 201 7. Tallinn: Eesti Keele Instituut.
Trosterud, T., Moshagen, S. & Pirinen, T. (2013). Building an open-source development infrastructure for language
technology projects. NEALT Proceedings Series, 16, pp. 343-352.
Völkel, M., Krötzsch, M., Vrandecic, D., Haller, H. & Studer, R. (2006). Semantic Wikipedia. Proceedings of the
15th international conference on World Wide Web (WWW ‘06) (pp. 585-594). ACM.
Viikberg, J. (2008). Eesti keele kogud. In E. Parmasto, & J. Viikberg, Eesti humanitaar- ja loodusteaduslikud
kogud, seisund, kasutamine, andmebaasid (pp. 95-112). Tartu: Tartu Ülikooli Kirjastus.