1. Could you briefly tell us about your academic background? What motivated you to apply a computational approach to linguistics?
I studied linguistics at Vytautas Magnus University (VMU). I was inspired to take up computational and corpus linguistics by Prof. Rūta Petrauskaitė, who is the founder of corpus linguistics in Lithuania and was my thesis advisor. The topic of my PhD was morphological ambiguity, which I analysed using a morphologically annotated corpus of Lithuanian. I defended my PhD in 2006 and am now a researcher at the Centre of Computational Linguistics and a lecturer at the Department of Lithuanian Studies at VMU.
I’ve been a member of the Centre of Computational Linguistics at VMU since my MA studies which has given me a lot of valuable opportunities to get involved in computational research. I was able to get acquainted with specialists in corpus and computational linguistics working in Lithuania and other countries, to try out different language analysis software and corpora, as well as to observe other developments in language research. The gained experience allowed me to specialize in automatic morphological analysis.
2. You’ve worked on quite a few important language projects with the Lithuanian CLARIN consortium. What is your contribution when collaborating with the consortium?
I’ve had the opportunity to contribute to the creation of some of the key resources in the CLARIN-LT infrastructure. For example, the first version of MATAS was compiled during my PhD studies to analyse the problem of morphological ambiguity, which had previously been only very scarcely investigated in Lithuania and abroad. Manual annotation of semi-automatically annotated texts helped me to describe this phenomenon in detail in my dissertation, which contributed to the development of more accurate automatic morphological annotation tools for Lithuanian. The revised version of MATAS was added to the CLARIN-LT repository so that it is now available for anyone interested in it.
3. Would you like to recommend a language resource or tool developed at the consortium that you think is important for the study and analysis of the Lithuanian language?
Since 2016, I’ve been leading the project Automatic Identification of Lithuanian Multi-word Expressions financed by the Research Council of Lithuania. The project aims to develop a methodology for analysing Lithuanian MWEs by creating or adapting necessary tools and resources. Apart from the MWE identification methodology, we also aim to create MWE extraction tools, a database of Lithuanian MWEs with multifunctional search options, and a corpus-based dictionary of Lithuanian collocations.
CLARIN-LT is a partner on the project, which to me is an example of a successful collaboration between CLARIN-LT and linguists. CLARIN-LT provides me with the technical support and create the tools necessary for the implementation of the project. In return for their support, we will upload all project results into the CLARIN-LT repository and make them easily accessible for other researchers. For example, at the end of the year, the first dictionary of Lithuanian collocations will be released. Users will be able to access a database of Lithuanian multiword units encompassing over 10,000 lemmas. I think that this is an important contribution both for the development of further lexicographic resources as well as language teaching, especially given that collocation dictionaries don’t yet exist for most under-resourced languages like Lithuanian.
4. You are also a teacher at the Department of Lithuanian Studies at the Vytautas Magnus University. How do you integrate the computational approach into your course-work? Do you introduce the CLARIN infrastructure to your students?
I cannot imagine my classes without introducing students to the morphologically and syntactically annotated corpora. Naturally, before starting work with the resources I introduce the main principles of CLARIN and the role of national repositories. I always encourage my students to use the Corpus of Contemporary Lithuanian Language, which was developed by CLARIN-LT, for example. Although corpora do not provide ready-made information in contrast with dictionaries, I believe it is important to teach students the importance of making linguistic claims on the basis of authentic language use. The students of BA and MA study programmes of Lithuanian Philology and Modern Linguistics where I teach, are taught to work with the Lithuanian Morphologically Annotated Corpus MATAS during the lectures on morphology and word formation. The students have to identify the missing node in the collocations extracted from the Corpus of Contemporary Lithuanian Language; identify parts of speech and grammatical categories in extracts from MATAS; analyse syntactic relations in ALKSNIS, etc. Apart from in-class activities, students also write seminar papers and BA and MA theses drawing on data extracted from the mentioned resources, some of them are even invited to work on our research projects. For example, Rūta Brinkutė’s MA thesis analyses the distribution of grammatical categories in different genres. I believe that knowing how to work with annotated corpora and tools might be valuable for students in their future work as language editors or researchers.
5. You have been part of the team that created the LILA corpus, which is a parallel corpus of Lithuanian and Latvian. The team included both Lithuanian and Latvian researchers involved with CLARIN. How does the Lithuanian CLARIN consortium benefit from such cross-border collaborations? Do you plan to upload the corpus into the consortium’s repository?
The project was part of the EU Cross-Border Cooperation Programme and was conducted in 2011-2012 before either Lithuania or Latvia were CLARIN members. The 9 million word Lithuanian-Latvian-Lithuanian parallel corpus aligned on paragraph and sentence level was compiled by researchers of the Vytautas Magnus University’s Centre of Computational Linguistics and the Latvian University’s Mathematical and Informatics Institute’s Laboratory of Artificial Intelligence (LU MII). It was a really interesting and mutually beneficial experience to work with the Latvian colleagues, as our teamwork not only resulted in the creation of the corpus itself, but also in several joint publications. I believe that if the project was implemented now, when both countries have CLARIN centres, the project aims and results could have been formulated on a much larger scale and more language pairs could have been included in the corpus. I see great value in such collaborative projects as they allow us to combine a wide variety of research perspectives and approaches, which in turn enhances professional and personal cooperation between the research centres and scientists in different countries.
6. What would you recommend CLARIN to do in order to attract more researchers from the Lithuanian linguistics community?
In relation to my previous comment on LILA corpus, I think that CLARIN could focus more on promoting joint scientific projects among the CLARIN centres of different countries to create comparable language resources and compatible processing tools. I also think that the fact that there is a consortium like CLARIN-LT which develops tools and resources specifically dedicated to Lithuanian can be very inspiring for new initiatives and research projects that also might want to start working with other less-resourced languages.
Also, I would like to see interoperable lexicographical databases to become available through CLARIN. At the very least, providing more information on the availability of such resources would be very helpful. For example, during the lexicographic project “Automatic Identification of Lithuanian Multi-word Expressions“, we were looking for a data base we could reuse for our research. As we found none, we spent a lot of time as well as human and financial resources to create the database ourselves.
Click here to read more about Tour de CLARIN