RAMUNĖ VASKELAITĖ

CHALLENGES POSED BY DIGITAL TECHNOLOGIES AND PROPOSED WAYS TO OVERCOME THEM

The 24th Jono Jablonskio Conference

Taken together, the Jono Jablonskis Conferences organized each autumn since 1993 by the Institute of the Lithuanian Language and the Department of Lithuanian Language at Vilnius University show that they have most often focused on fairly specific problems of standardization and usage, or on the relationship between the norm and usage (e.g., Language Norms and Usage, 1994; New Language Phenomena and Their Evaluation, 1998; Changing Norms and Usage, 1999; Levels of Language Norms and Usage, 2000; Language Usage and Management, 2001)¹. However, there has probably not been a single conference at which more fundamental questions did not arise in one way or another—questions concerning the concept and boundaries of the standard language, its purpose, its relationship with society, its adaptation as circumstances change, and its survival. And since around the turn of the century, as the various circumstances affecting language have undergone increasingly profound changes, such questions have also been deliberately foregrounded as the main subject of discussion (cf. Boundaries of the Standard Language, 2002; The Standard Language and Society, 2005; The Standard Language and Other Language Varieties, 2006; Language Variation and Linguistic Attitudes, 2009; The Life and Renewal of the Lithuanian Language in Contemporary Society, 2013).

The development of the conferences reveals another tendency: as attention turns to fundamental issues, thereby deepening in this respect, it also embraces a broader and broader perspective. Four conferences have already been held as international events², which means that they have considered not only the state or concept of the Lithuanian standard language, but also standard-language concepts developed in other countries or, where no such concepts are applied, directions in language planning, policy, and management.

This year’s 24th conference opened up a particularly broad perspective—that of Europe, encompassing the multilingual European Union and the problem of linguistic diversity arising within it. In terms of its subject matter, the conference held at the Institute of the Lithuanian Language, DIGITAL LANGUAGE RESOURCES, THEIR DEVELOPMENT DIRECTIONS AND USE POSSIBILITIES (September 29, 2017), could also be linked to an even broader scale—that of the global. And although Lithuanian words such as worldwide; universal; huge have been proposed as alternatives to the adjective global,

¹ Abstracts of papers presented at some J. Jablonskis Conferences can be found on the Institute of the Lithuanian Language’s website at http://lki.lt/tarptautin%C4%97-jono-jablonskio-konferencija/#1501063755178-1d439107-84e9. In the Standard Language (formerly Language Culture), conference reviews are published.

² International conferences were organized in 2005, 2009, 2010, and 2016.

BENDRINĖ KALBA 90 (2017)

www.bendrinekalba.lt

ISSN 2351-7204

RAMUNĖ VASKELAITĖ. Skaitmeninių technologijų keliami iššūkiai ir siūlomi būdai juos įveikti | 2

or enormous³, this year’s discussions are most accurately described by an international term encompassing several semantic nuances. First, the digital space is not only global, but also universal, enormous or even boundless, especially since the speakers cited not only computers as digital technologies, but also smartwatches, phones, voice-controlled vacuum cleaners, robots and artificial intelligence in general. Second, the word global immediately reveals links with globalization and globalizing processes, and such obvious links are found not only in the area of language use chosen for examination, but also in the fundamental question raised at the conference.

The question could be summed up by the formulation offered by Laimutis Telksnys, habilitated doctor of mathematical sciences: “How should we, a few million Lithuanians, act so that we do not disappear in the ocean of more than 7 billion people who write and speak, and of the smart machines to come?” Yet the concrete, matter-of-fact answer given by the academic at the plenary session, as well as the joint projects presented one after another by philologists, mathematicians, programmers and engineers at the conference’s other four sessions, showed that the general strategic direction was well known. It is the incorporation of language and its structure into the digital space, and active efforts to ensure that it functions there, both in written and spoken form. At the conference, this strategic direction was not so much debated as regarded as the only way for a small country’s language to survive in the digital age, the only possible choice. Member of the European Parliament Algirdas Saudargas compared it with the choice faced by the book smugglers: in an environment where Cyrillic was used, they chose Lithuanian writing in Latin letters. Dr Audrius Valotka, project manager at the Faculty of Philology of Vilnius University, spoke of a chain of successive choices: peoples who chose to record something in clay or on parchment survived; later, along the path of history, those peoples survived who printed texts on Gutenberg’s invented printing press; now it is the turn to make the inevitable choice to move into the digital space.

Difficulties encountered while moving in a clearly understood direction were mentioned somewhat more often at the conference than the strategic direction itself, but they were not the main focus of its discussions either. Several speakers identified as a problem the dominance in the world of technology of United States (hereinafter—the US) companies such as Microsoft, Amazon or Google, which are little interested in the markets of small countries, let alone their languages. This is why the conclusion was reached several times that we ourselves must take responsibility for the use and survival of our language in the digital space. However, more than one person already working on this issue mentioned a shortage of resources, including human resources. Understandably, a small country has little chance of competing in language technologies with the market’s sharks or giants—often identified as US private-sector companies—but the speaker who discussed the situation in the European Union (hereinafter—the EU)

³ For example, such substitutes are listed in the Consultation Bank provided by the State Commission of the Lithuanian Language: http://www.vlkk.lt/konsultacijos/5531-globalus.

BENDRINĖ KALBA 90 (2017) www.bendrinekalba.lt ISSN 2351-7204

RAMUNĖ VASKELAITĖ. Skaitmeninių technologijų keliami iššūkiai ir siūlomi būdai juos įveikti | 3

the presentation by Spanish scholar Rafael Rivera made clear that the dominance of these companies in the information technology world also poses problems at EU level. On the other hand, the EU itself is also drifting towards English, which should be a factor that makes the introduction and continued use of Lithuanian in the digital space especially difficult.

However, the conference opened up a completely different perspective. The plenary session began with a presentation entitled The Living Language in Artificial Intelligence by Algirdas Saudargas, signatory of the Act of Independence and member of the European Parliament. Prompted by a study conducted by META-NET, a network of excellence comprising 60 research centres in 34 countries, which showed that Lithuanian received the lowest score in terms of the state of language technologies—that is, it was assessed as “having little or no technological support,” meaning that the language was at risk of disappearing—he had initiated a study in the European Parliament several years earlier, Language Equality in the Digital Age. The study found that the condition for achieving genuine integration in the EU was not monolingualism but, rather, linguistic diversity. It was later supplemented by the Mother Tongue Project, which proposes strategic approaches at various levels for achieving a technological breakthrough in the EU that would enable every EU citizen to access services and communicate in the common digital space regardless of the language they use.

Rafael Rivera, director of the consulting firm iClaves and one of the researchers, came to present the study and the report prepared on its basis. He identified a contradiction arising from linguistic diversity in the EU: on the one hand, as an expression of Europe’s cultural diversity, it is considered one of the greatest assets of a union with 24 official languages and more than 60 national and regional minority languages; on the other, the language barriers that arise are among the main obstacles to creating a truly integrated union. These barriers particularly affect people with less education and older people, as well as speakers of smaller and regional minority languages. This results in linguistic exclusion. The economic and social consequences of language barriers are substantial: the opportunities offered by the single market, the free movement of goods and labour, public participation in political processes and electronic commerce are not fully exploited, and citizens entitled to use these opportunities cannot do so adequately. This problem becomes especially pressing in the digital age. New technologies, such as self-learning neural networks, are creating ways to overcome language barriers, but progress is hampered by the fragmentation of the European language technology market, uncoordinated research, insufficient funding, and the inadequate attention given to these issues on the EU political agenda. All these causes and problems explain

BENDRINĖ KALBA 90 (2017)

www.bendrinekalba.lt

ISSN 2351-7204

RAMUNĖ VASKELAITĖ. Skaitmeninių technologijų keliami iššūkiai ir siūlomi būdai juos įveikti | 4

the emergence of the European Mother Tongue Project, which envisages a comprehensive set of measures spanning institutional, scientific, industrial, market and public-sector levels⁴.

In his presentation Written and Spoken Lithuanian in Traditional and Electronic Environments⁵, Laimutis Telksnys, habilitated doctor, professor at the Institute of Mathematics and Informatics of Vilnius University, a pioneer of internet development in Lithuania and a scientist renowned for his inventions in Lithuania and beyond, answered his own question about the possibility of surviving in an ocean of billions by formulating quite concrete tasks as essential conditions for survival: to ensure that correct Lithuanian is preserved in paper documents and electronic environments, written and spoken language rules and technical and software tools are needed, as are transcribers that would allow characters from non-Lithuanian writing systems to be rendered using Lithuanian written-language characters, and spoken-language sounds to be represented by Lithuanian spoken-language symbols suitable for the automatic synthesis of the sounds of Lithuanian speech words. This would ensure that computers, phones, other speaking smart devices and robots could all be addressed in correct written and spoken Lithuanian, and that non-Lithuanian personal names in Lithuania’s electronic data repositories would be represented using Lithuanian character codes alongside their original spellings; their pronunciation would be provided by audio recordings made both by pronouncing them using the phonemes of Lithuanian and by recordings in the original language.

The professor himself and other specialists at the Institute of Mathematics and Informatics have already done much to ensure the use of correct spoken Lithuanian in the digital space, as became clear from the presentations delivered at the first session, SPEECH RECOGNITION AND SYNTHESIS. (Indeed, at all the conference’s other sessions, difficulties were mentioned only briefly; the work being done was presented in a practical and constructive manner, and further tasks were outlined.) In collaboration with philologists—the Institute of the Lithuanian Language, the Faculty of Philology of Vilnius University, Vilnius Pedagogical University and Šiauliai University—a group of specialists from Vilnius University’s Institute of Mathematics and Informatics and Faculty of Mathematics and Informatics, led by habilitated doctor L. Telksnys, implemented the project Lithuanian Speech-Controlled Services (LIEPA) in 2013–2015. The result was seven services, available to everyone and provided in correct spoken Lithuanian: Pažintuvas, which helps explain cell structure in schools; Ieškotuvas, which introduces users to UNESCO heritage sites; and applications that allow users to control a computer and conduct searches using voice commands

⁴ The study Language equality in the digital age. Towards a Human Language Project and its English summary are publicly available. They are available on the European Parliament website (http://www.europarl.europa.eu/thinktank/en/document.html?reference=EPRS_STU(2017)598621),, but the text may be reproduced or translated for non-commercial purposes only with the European Parliament’s consent.

⁵ For an article prepared on the basis of the presentation, see General Language in the “Reviews” section of this volume.

BENDRINĖ KALBA 90 (2017)

www.bendrinekalba.lt

ISSN 2351-7204

RAMUNĖ VASKELAITĖ. Skaitmeninių technologijų keliami iššūkiai ir siūlomi būdai juos įveikti | 5

Controller and Browser, Assistant and Tartuvas, which speak correct Lithuanian, as well as a computer synthesizer for blind people that reads text on a computer screen. Infrastructure services were also developed: an electronic text reader, a voice-operated program for opening internet pages, a Lithuanian speech-recognition engine, a speech corpus, and a voice-recording laboratory⁶. It might seem that much has been accomplished, yet L. Telksnys, while presenting a paper prepared with Gediminas Navickas, a senior specialist in the Projects Division of the same institute, entitled Voice-Controlled Services in Lithuanian. Situation. Prospects, described this work only as opening the door for Lithuanian spoken language to enter the electronic space. The presenter was inclined to cite as project outcomes not so much the services developed as the highlighted importance of “fusing” the knowledge held by computer scientists and philologists, and the value of setting an example—showing manufacturers and service providers the possibilities for using Lithuanian voice-controlled services. The scientist, who sees global trends in technology, considered a rather urgent task now to be the inclusion of robots in school curricula—a task falling not so much to language-technology developers and promoters—and, before that, the training of teachers.

The head of projects at Vilnius University, Dr. Audrius Valotka, likewise described the work on the project LIEPA as merely opening the gates. In the paper Beyond Gutenberg’s Printing Press: Lithuanian Speech in the Latest Technologies (co-authored with G. Navickas), he stated that “at present, Lithuanian speech in information technology is in a situation similar to that of 30 years ago, when questions concerning the use of Lithuanian diacritical marks in computing were being addressed.” This state of affairs calls for further work—the project LIEPA 2. The outcome of this project, already under way, should be infrastructure solutions: a 1,000-hour Lithuanian speech corpus, a Lithuanian speech recognizer with a 3,000-word vocabulary, and an improved Lithuanian speech synthesizer. New services will also be created: a humanoid robot that develops children’s decision-making abilities; a voice-controlled system for managing contacts on a mobile phone; a taxi-calling service; an internet news reader; and an interlingual communicator covering Lithuanian and Chinese. However, the most important outcome envisaged for the new project is the ability to communicate with objects (mobile phones, tablets, smartwatches, robots) in natural speech, give commands in a human voice, and understand their responses.

Another representative of the Institute of Mathematics and Informatics, Associate Professor Pijus Kasparaitis, in his paper The Present and Prospects of Computer Speech Synthesis in Lithuanian (co-authors: Dr. Gintaras Skersys, an assistant at the Faculty of Mathematics and Informatics of Vilnius University, and lecturer

⁶ They are described in greater detail on the website for Integrated Lithuanian Language and Writing Resources, Raštija.lt: https://liepa.rastija.lt.

BENDRINĖ KALBA 90 (2017)

www.bendrinekalba.lt

ISSN 2351-7204

RAMUNĖ VASKELAITĖ. Skaitmeninių technologijų keliami iššūkiai ir siūlomi būdai juos įveikti | 6

Margarita Beniuše)⁷, using video footage and websites, illustrated the potential uses of the synthesizer developed during the LIEPA project: the synthesizer was used in the performance Remote Vilnius, presented at the 2014 theater festival Sirenos (audience members, who also took on the actors’ roles, walked through the streets of Vilnius listening to synthetic voices through headphones); articles in some online newspapers (lzinios.lt, ukininkopatarejas.lt, vilnius.lt, delfi.lt) can not only be read but also listened to; announcements for passengers at the Vilkaviškis bus station are read aloud in a synthetic voice; the text-to-speech service RoboBraille automatically converts any text documents into audio files; a Lithuanian-speaking robot has also become a reality—the humanoid NAO invited visitors to see Lithuania’s stand at an event presenting EU countries held in Brussels, and recited Kristijonas Donelaitis’s Years at the conference Will Robots Speak Lithuanian?, held at the Seimas. Listing future applications of speech synthesis, some steps toward which have already been taken, P. Kasparaitis mentioned audio versions of articles on Linux-based websites, virtual assistants, and a mobile vending machine that could offer snacks or drinks in a park, the MIO mobile kiosk.

The work presented in the next three sessions, largely concerned with efforts to preserve written language, gave the impression that the digitization of dictionaries, text corpora, databases, and card indexes is increasingly giving way to a form of language digitization that would enable language resources to be used not only by language users and researchers, but also by language-technology programs and artificial intelligence. This strategic direction, increasingly relevant as artificial intelligence makes ever greater advances, was emphasized by language strategist A. Saudargas both in press articles⁸ and at this conference. In the abstracts submitted to the conference, it was described as follows: “The development of artificial intelligence is converging toward a hybrid form, in which neural networks correspond to the brain’s unconscious mechanisms, while conscious mental activity is modeled by traditional, so-called symbolic artificial intelligence. The same applies to language technologies if we understand them as models of the mechanisms of language in the human brain (living language). Therefore, every linguistic community must ensure that language technologies corresponding to symbolic artificial intelligence (so-called ‘rule-based’ programs) reflect the entire structure of the native language comprehensively and accurately, and that a rich environment of the native language (and native culture) is created for neural networks to learn by themselves.”

⁷ For an article based on the presentation, see the “Reviews” section of this volume of Bendrinės kalbos.

⁸ For example, in the article Fulgura plango, published in the daily Bernardinai.lt (http://www.bernardinai.lt/straipsnis/2012-12-05-algirdas-saudargas-fulgura-plango/91787),, in the article Native Language: What Google Won’t Solve, published in Lietuvos žinios (http://lzinios.lt/lzinios/Mokslas-ir-svietimas/gimtoji-kalba-ko-google-neisspres/246316).

BENDRINĖ KALBA 90 (2017)

www.bendrinekalba.lt

ISSN 2351-7204

RAMUNĖ VASKELAITĖ. Skaitmeninių technologijų keliami iššūkiai ir siūlomi būdai juos įveikti | 7

At the conference’s second session, DIGITAL RESOURCES OF THE LITHUANIAN LANGUAGE, which presented a new project being carried out by the Institute of the Lithuanian Language, it was likewise stated that as new needs, challenges, and expectations arise, language resources must change: new ideologies and concepts for digital language resources must emerge, corridors of meaning and concept engineering must be developed, a variety of information retrieval options must be created, and so on. In the presentation read by Deimantė Budriūnaitė, head of the General Affairs Infrastructure Division, E. kalba – skaitmeninių kalbos išteklių naudojimo(si) inovacija, the Institute’s new project was introduced. She is carrying it out together with director Jolanta Zabarskaitė and Netcode project manager Skirmantas Šermukšnis. In fact, it is a continuation of an earlier project on language digitization. At that time, 11 monolingual and bilingual dictionaries and 5 card files were digitized, thereby creating the Information System of Lithuanian Language Resources⁹. The new project aims not only to integrate new resources, but also to develop technologies, solutions, and tools for semantic searches of resources, visualize their results, and use these technologies to create services for users: Search in the Word Network, E-marketing, E-concepts, and E-advice. The planned capabilities include adapting resources for marketing purposes through the provision of public and private electronic services, ensuring wider online availability of word-formation tools and language advice, and creating opportunities to link the Lithuanian word network with the word networks of more than 200 other languages.

Concern for the use of Lithuanian language resources already made available online was also reflected in two databases presented by representatives of the Institute of the Lithuanian Language. One has been integrated into the Information System of Lithuanian Language Resources, while the other is to be integrated. Dr. Rita Miliūnaitė, initiator and creator of the Database of Lithuanian Neologisms and senior researcher at the Institute of the Lithuanian Language’s Centre for Standard Language Research, described the database she is developing as an online reference work intended to record and lexicographically describe Lithuanian neologisms from the early 21st century¹⁰. Her presentation, however, revealed that at the current stage of its development, the database’s usefulness to users is perhaps the principal concern. To this end, searches have been enabled by headword, neologism origin, original form, spelling variants, field of use, certain distinctive features, and the surname of a prominent person who used the neologism or its author; an advanced search system is also being developed for specific user groups. Interest in both the database and neologisms is high, so far the only problem mentioned being that the data submitted by users cannot be processed quickly enough. On the other hand, the large number of submissions debunks two persistent myths in society: that the Lithuanian language is dying and that linguists create words. Nevertheless

⁹ Information System of Lithuanian Language Resources is available at http://lkiis.lki.lt/home.

¹⁰ The Database of Lithuanian Neologisms can be accessed at http://naujazodziai.lkli.lt.

BENDRINĖ KALBA 90 (2017)

www.bendrinekalba.lt

ISSN 2351-7204

RAMUNĖ VASKELAITĖ. Skaitmeninių technologijų keliami iššūkiai ir siūlomi būdai juos įveikti | 8

Dr. Daiva Murmulaitytė, who also works at the Centre for Standard Language Research, suggested in her presentation, Prospects for Research into Neologism Formation (The Case of the Database of Lithuanian Neologisms), that steps also be taken to make the database more useful to language researchers. This would require adapting it accordingly. It was proposed that neologisms be marked with indexes recording the part of speech of the base word, changes in the derivational base, the role of analogy and translation in creating a neologism, instances of atypical word formation, and so on. The speaker noted that attempts to do this had revealed word-formation phenomena in neologisms that have been little studied or not studied at all.

Listing the objectives of another database—the Lithuanian Place Names Geoinformation Database¹¹—Dr. Laimutis Bilkis, senior researcher at the Institute of the Lithuanian Language’s Centre for Baltic Languages and Onomastic Research, also mentioned both practical aims, namely teaching correct usage, and linguistic research (including the preservation and dissemination of language). That the database is intended for both language users and researchers is evident from its search and information-display options: information about place names can be searched by name, gender, number, stress pattern, word-formation method, origin according to the language of the base word, and origin according to the lexical group to which the base word belongs. The place-name analysis section provides stress patterns for all inflected forms, as well as explanations of origin and word formation. About 25,000 place names have been uploaded to the database. Some links have also been established: integrating the database into the Information System of Lithuanian Language Resources created links to another place name from which the one being searched for is derived; to a personal name in the Lithuanian Surname Database from which the place name is derived; and to historical spellings of that place name in the Historical Place Names Database. Links to standard and terminological vocabulary are also planned, especially semantic, derivational, and locative links. Several problems were mentioned, however—not only the sheer number of place names, but also fragmented funding.

A clear shift from collecting language data to exploring how to use it is also reflected in the work being done by representatives of corpus linguistics. At the conference’s third session, CORPUS LINGUISTICS, two collaborative projects were presented by researchers from Vytautas Magnus University’s Centre of Computational Linguistics, who do most of this work in Lithuania, together with representatives of the Baltic Institute of Advanced Technology. The projects demonstrate an effort to annotate corpora and thereby use them for linguistic research and for creating other language resources. The presentation prepared by the first group of corpus linguistics researchers—Erika Rimkutė, Agnė Bielinskienė, Loïc Boizou, and Andrius Utka—¹²introduced the morphologically annotated corpus MATAS and the syntactically annotated corpus ALKSNIS. The MATAS corpus was prepared on the basis of a corpus of 1 million words,

¹¹ This database is also available at http://lkiis.lki.lt/lietuvos-vietovardziu-geoinformacine-duomenu-baze.

¹² See the article based on the presentation in the “Language Technologies” section of Bendrinės kalbos in this volume.

BENDRINĖ KALBA 90 (2017)

www.bendrinekalba.lt

ISSN 2351-7204

RAMUNĖ VASKELAITĖ. Skaitmeninių technologijų keliami iššūkiai ir siūlomi būdai juos įveikti | 9

After statistical models were applied, it was annotated using a morphological annotator developed by the Centre of Computational Linguistics. ALKSNIO annotation is based on the principles of automatic morphological and syntactic annotation, employing a syntactic dependency model. The project carried out by the second group of researchers—Erika Rimkutė, Agnė Bielinskienė, Loïc Boizou, Ieva Bumbulienė, Jolanta Kovalevskaitė, Tomas Krilavičius, Justina Mandravickaitė, and Laura Vilkaitė—revealed that, after annotating the Delfi.lt corpus, work is moving toward developing a methodology for studying fixed word combinations in contemporary written Lithuanian and compiling a Lithuanian collocations dictionary. The database of fixed word combinations being compiled on the basis of the corpus is intended to provide grammatical and lexical information about the combinations, their frequency of use, text section, concordance examples, and so forth. The database currently contains about 2,700 lemmas and more than 35,000 inflected forms.

The practical possibilities for using corpora were also demonstrated in a presentation prepared by Gintarė Judžentytė and Dr. Vilma Zubaitienė, lecturers in the Department of Lithuanian Language at Vilnius University, Corpus-Based Studies of Academic Phrases: Formal Structure and Semantics. After reviewing previous attempts to study the language of science on the basis of corpora, they also presented their own ongoing project, which aims to investigate the structure and semantics of academic-language phrases using the Student Writing Corpus being developed for this purpose at Vilnius University. Drawing on the corpus material and the results of the study, a collection of academic-language phrases will be compiled, useful to students in various degree programs and fields of study.

The fourth conference session, AUTOMATED ANALYSIS OF LANGUAGE STRUCTURE AND TEXTS, revealed even more sophisticated possibilities for using new technologies to analyze language. In his presentation, Lithuanian Grammar in the Digital Open-Source World¹³, engineer Virginijus Dadurkevičius of Vilnius University’s Institute of Applied Sciences reviewed the work in morphological analysis carried out by Vytautas Zinkevičius, a programmer at the Institute of the Lithuanian Language, which for nearly two decades had been an indispensable and sole tool for modern Lithuanian-language research and computational applications. He also presented a certain shift in computational morphology made possible by the use of the Hunspell platform. The language’s morphology is described on the platform in two text files, both based on the Grammar of Contemporary Lithuanian and the corpora of Vilnius University, Vytautas Magnus University, and the Seimas of the Republic of Lithuania. According to the presenter, approximately 5,000 groups of addressable rules have been recorded in total, and more than 170,000 lemmas selected. In selecting the lemmas, only standard contemporary Lithuanian was included. According to the presenter, this morphological analyzer correctly

¹³ See the article based on the presentation in the “Language Technologies” section of this volume of Bendrinė kalba.

BENDRINĖ KALBA 90 (2017)

www.bendrinekalba.lt

ISSN 2351-7204

RAMUNĖ VASKELAITĖ. Skaitmeninių technologijų keliami iššūkiai ir siūlomi būdai juos įveikti | 10

interprets an average of 99 out of every 100 words. The new method of morphological analysis was successfully applied in the Lithuanian Language Syntactic and Semantic Analysis Information System¹⁴ developed by Vytautas Magnus University, in the Register of Legal Acts¹⁵, and with particular success in the Solr/Lucene document indexing and search system (according to the presenter, it proved possible not only to index and find all the desired word forms accurately, but also to speed up the indexing process hundreds of times). A specific next step is considered to be the preparation of two dictionary versions: a normative version suitable for spell-checking, and a version containing both normative and non-normative language, suitable for more applied tasks in the analysis of spoken language.

The opportunities opened up by the digital era for research into old writings were described by Dr. Mindaugas Šinkūnas, an employee of the Institute of the Lithuanian Language’s Centre for Research into Written Heritage, in his presentation Who Copied from Whom? Automating and Visualizing the Study of the History of Bible Translations. He presented a methodology that has already been tested and makes it easier to assess the similarity of quotations in Bible translations—a task that poses the greatest difficulties for researchers of old writings. According to the presenter, the methodology’s promise is demonstrated by the fact that the results of a software-assisted comparison of biblical verses in Bretkūnas’s Postilė (1591) correspond to the conclusions reached by earlier researchers using other methods. The presenter considered the most important outcome of the work to be a tool that indicates the approximate degree of similarity between sources or parts of them. It opens the way to further, more detailed research.

Naturally, the conference also addressed an area of language technology that has experienced many failures—automated translation—not least because the term language technologies is applied primarily to it. In his presentation, Machine Translation for Lithuanian¹⁶, engineer Danielius Algirdas Ralys of Vilnius University’s Institute of Applied Sciences reviewed, stage by stage, all the achievements and failures of machine translation (the literal translation that emerged during the Cold War, rule-based translation, and statistical translation methods), and concluded that Lithuania has not been left behind in this field. With the emergence of neural networks that construct transformation rules, hopes for a new breakthrough in automated translation are pinned on them. Noting that the latest neural machine translation systems translate better than an average translator, he presented Vilnius University’s plans to launch a project in 2018 on next-generation neural translation in English, Lithuanian, Polish, French, Russian, and German. Nevertheless, it was observed that efforts should focus not only on copying what has already been invented, but also on creating something original and new.

A project underway at the Institute of the Lithuanian Language revealed that such efforts are indeed being made. That it is important for Lithuania to develop alternative machine-translation options was emphasized by the Institute of the Lithuanian Language’s

¹⁴ address of the Lithuanian Language Syntactic and Semantic Analysis Information System – https://semantika.lt.

¹⁵ address of the Register of Legal Acts – https://www.e-tar.lt/portal/index.html.

¹⁶ See the article based on the presentation in the “Language Technologies” section of this volume of Bendrinė kalba.

BENDRINĖ KALBA 90 (2017)

www.bendrinekalba.lt

ISSN 2351-7204

RAMUNĖ VASKELAITĖ. Skaitmeninių technologijų keliami iššūkiai ir siūlomi būdai juos įveikti | 11

Dr Daiva Šveikauskienė, a researcher at the Centre for Standard Lithuanian Studies, and programmer Vytautas Šveikauskas decided to draw on 2017 Tildės data, which show that the quality of translations into and out of Lithuanian produced by Google translation systems using statistical methods is poorer than that for Latvian or Estonian. The authors link the prospects for alternative translation options to the digital grammar of Lithuanian that they are developing. According to the presenters, so far only a pilot example of the grammar has been prepared, covering just a few words, but even this shows that translation quality is much better. The digital grammar, consisting of two levels—the abstract (forms common to all languages) and the concrete (a level reflecting the specific features of a particular language)—makes it possible to overcome what is perhaps the central problem of machine translation: conveying meaning. During translation, it allows a sentence to be transformed into an abstract semantic form, from which a sentence in another language is generated based solely on the properties of the target language, without taking into account the morphology of the language from which it was translated. This makes it possible to achieve far more accurate results in conveying meaning. Incidentally, the current translations into Lithuanian provided by Google were mentioned by nearly every other conference participant—as a failure of language technology, as a threat to the language, and as an example clearly demonstrating that we must take care of language technologies ourselves.

Summing up this year’s Jonas Jablonskis conference, several fairly interrelated conclusions naturally arise. First, the age of new technologies opens up extraordinary possibilities for language research: corpora and other collections of computerised data have revolutionised linguistics, while technologies offer ever more advanced possibilities for searching, summarising, and presenting statistics, further enhanced by the development of artificial intelligence. Yet new technologies pose particularly great challenges to the vitality of language, and the threat concerns not only standard Lithuanian but the Lithuanian language as a whole. Nevertheless, it is these technologies that offer ways to overcome the challenges, and the new European initiative creates especially favourable opportunities for the language of a small country to survive. At the same time, both this initiative and the development of digital technologies highlight an unprecedented need to pool and synthesise efforts. The projects presented at this conference already showed that, under such conditions, ensuring the survival of the language and researching it are tasks that language specialists can no longer accomplish using their knowledge and resources alone: philologists carried out projects together with computer scientists, mathematicians, and programmers. As A. Saudargas has pointed out more than once, the development of artificial intelligence creates a need not only to pool the efforts of specialists from different fields, but also to develop an interdisciplinary science taking shape at the intersection of language, computing, and cognitive sciences¹⁷. And the EU

¹⁷ For example, in the article The Lithuanian Language Must Become Literate in the Digital Age, published in the online daily Bernardinai.lt (http://www.bernardinai.lt/straipsnis/2013-01-25-algirdas-saudargas-lietuviu-kalba-privalo-tapti-rastinga-ir-skaitmenineje-epochoje/94233).

BENDRINĖ KALBA 90 (2017)

www.bendrinekalba.lt

ISSN 2351-7204

RAMUNĖ VASKELAITĖ. Skaitmeninių technologijų keliami iššūkiai ir siūlomi būdai juos įveikti | 12

the native-language project also reveals the necessity of pooling the efforts of the entire state—those of both the public and private sectors, politicians, researchers, and technologists.

Only at first glance might this year’s conference theme have seemed to have little to do with the name of the creator of standard Lithuanian. Anticipating the impression that there was no connection, one of the conference organisers, Dr Rita Miliūnaitė, identified one such link in her opening remarks: today, as language is brought into the digital space, just as in J. Jablonskis’s time, the aim is to create a universal, convenient form of language that meets society’s needs. Other conference participants also articulated one or another connection: P. Kasparaitis, expressing hope that the LIEPA project would become synonymous with caring for the Lithuanian language, added that J. Jablonskis would have approved; and D. A. Ralys recalled the creator of standard Lithuanian as a pioneer in formulating grammatical rules, a path from which work proceeded to automated morphological analysis. But the conference itself revealed far more shared connections and aims, above all a concern for the survival of the Lithuanian language and the use of correct language, as the world and the circumstances affecting language change at an astonishing pace.

Received 2017 10 22

RAMUNĖ VASKELAITĖ

[email protected]

BENDRINĖ KALBA 90 (2017) www.bendrinekalba.lt ISSN 2351-7204