Search

It is well known that AI-based language technology—large language models, machine translation systems, multilingual dictionaries, and corpora—is currently limited to three percent of the world’s most widely spoken, financially and politically backed languages. In response, recent efforts have sought to address the “digital language divide” by extending the reach of large language models to “underserved languages.” We show how some of these efforts tend to produce flawed solutions that adhere to a hard-wired representational preference for certain languages, which we call language modeling bias. Language modeling bias is a specific and under-studied form of linguistic bias were language technology by design favors certain languages, dialects, or sociolects with respect to others. We show that language modeling bias can result in systems that, while being precise regarding languages and cultures of dominant powers, are limited in the expression of socio-culturally relevant notions of other communities. We further argue that at the root of this problem lies a systematic tendency of technology developer communities to apply a simplistic understanding of diversity which does not do justice to the more profound differences that languages, and ultimately the communities that speak them, embody. Drawing on the concept of epistemic injustice, we point to the broader ethico-political implications and show how it can lead not only to a disregard for valuable aspects of diversity but also to an under-representation of the needs of marginalized language communities. Finally, we present an alternative socio-technical approach that is designed to tackle some of the analyzed problems. [ABSTRACT FROM AUTHOR]

adaptNMT streamlines all processes involved in the development and deployment of RNN and Transformer neural translation models. As an open-source application, it is designed for both technical and non-technical users who work in the field of machine translation. Built upon the widely-adopted OpenNMT ecosystem, the application is particularly useful for new entrants to the field since the setup of the development environment and creation of train, validation and test splits is greatly simplified. Graphing, embedded within the application, illustrates the progress of model training, and SentencePiece is used for creating subword segmentation models. Hyperparameter customization is facilitated through an intuitive user interface, and a single-click model development approach has been implemented. Models developed by adaptNMT can be evaluated using a range of metrics, and deployed as a translation service within the application. To support eco-friendly research in the NLP space, a green report also flags the power consumption and kgCO 2 emissions generated during model development. The application is freely available (http://github.com/adaptNMT). [ABSTRACT FROM AUTHOR]

The convergence of linguistics and computer science has given rise to the interdisciplinary field of Computational Linguistics, a domain dedicated to developing algorithms and models that facilitate the interaction between computers and human language. This abstract explores the fundamental components of Computational Linguistics, including Natural Language Processing (NLP), machine translation, speech recognition, and information retrieval. The applications of Computational Linguistics span diverse sectors, from chatbots and virtual assistants to sentiment analysis and healthcare informatics. Despite challenges in handling linguistic nuances, the field continues to advance with the integration of deep learning and neural networks. The collaboration between linguists and computer scientists is essential for navigating the complexities of language, as Computational Linguistics remains pivotal in shaping the future of digital communication and technology. [ABSTRACT FROM AUTHOR]

Copyright of Internationale Neerlandistiek is the property of Amsterdam University Press and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

The paper focuses on the use case of parliamentary debates as part of Digital Humanities. First, the ParlaMint project is outlined as a flagship initiative of CLARIN ERIC infrastructure. The project makes content from the national and regional parliaments visible, comparable and accessible for policy making and research. Then, the approaches are considered that have been applied in the creation of 31 corpora from national and regional parliaments. Last but not least, the utility of the multilingual resource is discussed.

Copyright of Fordítástudomány is the property of Faculty of Humanities ELTE, Department of Translations & Interpreting and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Artiklis uurime adjektiivi morfosüntaktilisi tunnuseid ja selgitame, kuivõrd on prototüüpsele adjektiivile omistatavad tunnused (nt ühildumine, võrrete moodustamine) adjektiiviklassile eriomased. Loome neile tunnustele tuginedes parameetrid, mille abil eristame korpuse andmete põhjal adjektiive teistest sõnaliikidest. Tüüpilise adjektiivi korpusprofiili tuvastamise kaugem eesmärk on rakenduse loomine, mis võimaldaks leksikograafidel ebaselgete juhtumite puhul kontrollida sõna adjektiviseerumise astet. Tutvustame kuue parameetri testimise tulemust 12 sõnarühma peal, millest igaühte kuulub 10 sõna. Sõnavalikul arvestame adjektiiviklassi piiripealseid juhtumeid ja leksikograafilisi kitsaskohti. Analüüsime, mil määral hälbivad erinevad testrühmad testitud parameetrite põhjal prototüüpsest adjektiiviklassi esindajast ning vaatleme ka variatsiooni adjektiiviklassi sees. Kõrvale kaldeanalüüs võimaldab välja selgitada parima eristusvõimega parameetrid. Eukleidilise kauguse mõõtmine eristab hästi adjektiivisarnased sõnad ja rühmad nendest, mis sarnanevad prototüüpsele adjektiivile vähem. *** How to recognize adjectives? An analysis of corpus patterns This study was inspired by a survey of Estonian lexicographers (Paulsen, Vainik and Tuulik 2019), where the lexicographers expressed the need for a new digital tool that would facilitate word class identification for ambiguous cases. In the case of adjectives, the lexicographers emphasized the difficulty of determining if a verb participle has sufficient adjectival use to be included in dictionaries as an adjective. In the article, we examine the morphosyntactic features characteristic of the adjective class and test different parameters in the corpora to differentiate adjectives from other word classes. We provide an overview of the test results of six parameters. In the study we analysed 12 groups of 10 words each. The test groups and test words were chosen manually, with consideration given to the problematic cases outlined by the lexicographers. We compared different types of adjectives or near to adjectives (the test groups) as well as different word classes (the control groups). To analyse the parameters’ capability to set adjectives apart, a deviation study was conducted. We determined a normative range for prototypical adjectives and set the minimum and maximum value for every parameter. In addition, we calculated the deviation of other test groups from the prototypical adjective range. The groups of particular focus (regular verb participles vs. adjectives) were best differentiated by three parameters. The sentence beginning testword+noun parameter (which determined if and how often a test word starts a sentence in the complement position) sets participles apart with 90% accuracy. Also, the parameter that measured the existence of comparative forms for test words was 100% accurate. The adverb parameter (which measured how often a test word is preceded by an adverb) was able to distinguish adjectives from verb participles with 80% accuracy. Among all groups, the comparative form parameter was the most accurate in the deviation study at setting prototypical adjectives apart from other test groups. A Euclidean distance analysis was able to differentiate adjective-like test words and test groups from others that do not behave similarly to prototypical adjectives. As all tested parameters produced meaningful results and were able to differentiate some word classes from adjectives, they can be input for a new digital tool which would show a word’s deviation from prototypical word class representatives to help lexicographers with word-class-related decisions.

The increasing use of language technology applications requires a more critical evaluation of the current state of language technology and its application than simply viewing it as an ideal and effective language learning aid. While an increased number of scholars have examined the issue of potential biases and hidden ideologies in language technology such as racism and gender discrimination, little attention has been paid to how the newly emerging language technology can contribute to reproduce the native speaker fallacy. This paper, focusing on the case of voice chatbots in Korea, critically examines how learning technology, in particular language technology applications, can potentially reproduce and reinforce the essentialist discourse of native speakerism, which posits native speaker accents as an ideal form of English and marginalizes nonnative English teachers and students. [ABSTRACT FROM AUTHOR]

In this paper, based on the example of our early works for Polish, we want to share our experience in the challenging task of developing NLP-based technologies in the situation of initial scarcity of digital language resources that ranked Polish among the Less-Resourced Languages. We present some of our projects aiming at language resources and tools we had to create in order to be able to process texts in Polish and develop real-scale systems with language understanding competence. The case study we present here is the rule-based system POLINT-112-SMS for improving information management in emergency situations. We argue in favour of the lexicon-grammar approach to the formal description of inflecting languages and present our current work on this grammatical paradigm. Our current work is on the implementation of the ideas presented in the first part of the paper on three prominent Indian languages, that is, Hindi, Odia, and Bengali.

During the COVID-19 pandemic, large amounts of data have been generated in social media, for the processing of which automatic methods such as topic modeling become necessary. In various studies, this has already provided fundamental insights into the topics discussed in different geographical regions and at different points in time. Other parameters such as authorship have also taken into account for additional differentiation, or individual sub-areas of corona discourse have been considered separately. Another interesting subfield here is science communication, whose importance for successful pandemic response is often mentioned. Here, too, topic modeling can serve as an approach in the future to examine science communication in the COVID-19 pandemic and its effects in more detail. [ABSTRACT FROM AUTHOR]

Artikkel tegeleb nimisõnavormide iseseisvumise küsimusega leksikograafia vajadustest lähtudes. Eeldusel, et abstraktseid käändevorme iseloomustab korpuses üldine püsiv esiletuleku proportsioon, pakume välja statistilise mõõdiku – distributsiooniindeksi, mille abil otsustada, kas sõnavormi kasutussagedus on piisav selleks, et lugeda teda paradigmast emantsipeerunuks ning seega iseseisva märksõna kandidaadiks. Indeks arvestab vormi suhtelist sagedust korpuses, võrreldes tegelikku ja normi põhjal oodatavat kasutussagedust, ning laseb samale skaalale paigutada väga erineva absoluutsagedusega juhtumeid. Artiklis illustreerime distributsiooniindeksi toimivust tavaliste rikkaliku vormistikuga nimisõnade ning ambivorme andvate sõnade vormistike võrdlusena. Seame provisoorse indeksi lävendväärtuse, millest suurema väärtusega vormi võib pidada iseseisvaks lekseemiks. Indeksit ning lävendväärtust testitakse erinevate korpuste (EtTenTen13, ÜK 2019) andmete peal. *** From inflected form to a word: The role of frequency This study is motivated by the need for a statistical benchmark that would help the lexicographer to judge a morphological form for its grammaticalization stage to the degree of an independent lexeme. The focus of this article is on Estonian substantives and in particular their forms in the 11 semantic cases. The choice of selection is based on the observation that the noun has a special position among the word classes with fuzzy categorial borders (Vainik et al. 2020) – the means of nominal morphology function as a source for ongoing processes of grammaticalization in Estonian. The case forms typically yield adverbs and adpositions – a phenomenon forming a part of ambiforms (words or forms that can be interpreted to belong to more than one word class). The main research question of this study is: is there a statistical sign indicating that a case form of a noun is emerging as a potentially independent lexeme? Based on the normal distribution of nominal case form frequencies, we established a statistic that determines a case form’s elicitation in a corpus – the distribution index (D-index). The D-index can be used as an indicator of the correspondence of a particular form’s actual frequency with the predicted elicitation degree. The D-index was tested by a sample of ambiforms (N = 46) and “ordinary” nouns (N = 26), the last group including nouns that display an abundant range of semantic case forms. This sample was used to extract all semantic case forms from altogether three Estonian corpora: the Balanced Corpus of Estonian, the National Corpus of Estonian 2019, and the web corpus etTenTen13. Based on the analysis, we defined a threshold value (≥ 0,130), indicating that the forms with higher D-indexes than this value can be regarded as independent lexemes. We conclude that the threshold value functions as a benchmark to a certain degree: an ambiform with a D-index over the threshold value is a distinctly independent lexeme. The forms with D-indexes below the threshold value may or may not be candidates of a lexical entry in a dictionary – the statistical parameters are not sufficient to make a waterproof decision. A lexicographer’s qualitative analysis will be needed in those cases.

Copyright of Eesti Rakenduslingvistika Ühingu Aastaraamat is the property of Eesti Rakenduslingvistika Uhing (ERU) / Estonian Association for Applied Linguistics and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

In this paper, based on the example of our early works for Polish, we want to share our experience in the challenging task of developing NLP-based technologies in the situation of initial scarcity of digital language resources that ranked Polish among the Less-Resourced Languages. We present some of our projects aiming at language resources and tools we had to create in order to be able to process texts in Polish and develop real-scale systems with language understanding competence. The case study we present here is the rule-based system POLINT-112-SMS for improving information management in emergency situations. We argue in favour of the lexicon-grammar approach to the formal description of inflecting languages and present our current work on this grammatical paradigm. Our current work is on the implementation of the ideas presented in the first part of the paper on three prominent Indian languages, that is, Hindi, Odia, and Bengali. [ABSTRACT FROM AUTHOR]

Artikli eesmärgiks on selgitada, millises ulatuses mõjutab keeleandmetele kohalduv õiguslik režiim keelemudelite arendamist ja kasutamist. Autorid lähtuvad oma käsitluses protsessiskeemist, alustades algandmetest ning lõpetades keeletehnoloogiat sisaldavate valmistoodetega (nt kõneliidesega külmik). Keeletehnoloogias kasutatavad algandmed sisaldavad tihti autoriõiguslikult kaitstavaid teoseid, autoriõigusega kaasnevate õiguste objekte (esitus, salvestus) ja isikuandmeid (isiku hääl, isiku kohta käiv muu info), mida säilitatakse annoteerimata ja annoteeritud andmekogudes. Keelandmete õiguslikke küsimusi on juba varem uuritud. Õiguslikult on läbi uurimata aga keelemudelite õiguslikud aspektid. Autorid on seisukohal, et reeglina ei mõjuta keelemudelite edasist õiguslikku staatust kasutatud algandmete õiguslik režiim, sest autoriõigusega kaitstavad teosed mudelis pigem ei säili. Küll aga võib õiguslikke probleeme tekitada isiku hääle kasutamine keelemudelis. Autorid analüüsivad võimalikke lahendusvariante nende probleemide ületamiseks. Artiklis vaadeldakse ka uue autoriõiguse direktiiviga kehtestatavat andmekaeve regulatsiooni ja selle rakendamist keelemudelite loomise kontekstis. *** "Influence of legal regime of language data on language models" This article aims to explain the extent to which the legal regime applicable to language data affects the development and use of language models. In their approach, the authors follow a process chart, starting from raw data to finished products containing language technology (eg a refrigerator with a speech interface). The raw data used in language technologies often include copyrighted works, objects of related rights (performances, sound recordings) and personal data (voice, other information about the person) stored in non-annotated and annotated databases. The legal issues of language data have already been studied. However, the legal aspects of language models have not been throughly explored. The authors are of the opinion that, as a rule, the legal status of the language models is not affect by the legal status of the used raw language data, since copyrighted works usually do not remain in the model. However, the use of a person’s voice in a language model can create legal problems. The authors analyze possible solutions to overcome these problems. The article also outlines the regulation of data mining introduced by the new copyright directive and its implementation in the context of development of language models.

In this paper we address how Natural Language Processing (NLP) approaches and language technology can contribute to data services in different ways; from providing social science users with new approaches and tools to explore oral and textual data, to enhancing the search, findability and retrieval of data sources. By using linguistic approaches we are able to process data, for example using Automated Speech Recognition (ASR) and named entity recognizers (NER), extract key concepts and terms, and improve search strategies. We provide examples of how computational linguistics contribute to and facilitate the mining and analysis of oral or textual material, for example (transcribed) interviews or oral histories, and show how free open source (OS) tools can be used very easily to gain a quick overview of the key features of text, which can be further exploited as useful metadata.

In this paper we address how Natural Language Processing (NLP) approaches and language technology can contribute to data services in different ways; from providing social science users with new approaches and tools to explore oral and textual data, to enhancing the search, findability and retrieval of data sources. By using linguistic approaches we are able to process data, for example using Automated Speech Recognition (ASR) and named entity recognizers (NER), extract key concepts and terms, and improve search strategies. We provide examples of how computational linguistics contribute to and facilitate the mining and analysis of oral or textual material, for example (transcribed) interviews or oral histories, and show how free open source (OS) tools can be used very easily to gain a quick overview of the key features of text, which can be further exploited as useful metadata. [ABSTRACT FROM AUTHOR]

Copyright of Eesti Rakenduslingvistika Ühingu Aastaraamat is the property of Eesti Rakenduslingvistika Uhing (ERU) / Estonian Association for Applied Linguistics and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

In this article we want to explore the status of the Dutch language. Is Dutch an important language, a 'major' one? How do we measure the importance of a language? Is this about the number of native speakers? Or are there other parameters that have to be taken into account? We can check the global influence of a language, the number of translations made, and the presence on the internet, amongst other elements. Ethnologue and UNESCO have developed a classification system to measure the importance and vitality of a language. Whether languages continue to be dynamic is also related to their digital power. In this article we will expand on the status of Dutch and the power of the Dutch digital language infrastructure. [ABSTRACT FROM AUTHOR]

Language technology provides several possibilities to commercialise collected language resources (data) in the form of providing access to databases, dictionaries, translation, text analysis and localisation services, storage of documents and personal language data, software, and other types of digital content. This article focuses on the contractual relationships in selling language technology products and providing services to the customers, and explores the entrepreneurial perspective on the supply of digital content to compare its compatibility with the existing regulatory framework. For comparison, the authors have chosen Estonian companies specialising in language technology. The main conclusions concerning the entrepreneurial perspective are based on a study of homepages and personal interviews. The compatibility between the business model and the regulatory framework is analysed based on the case study and Estonian legislation. The authors aim to outline how language technology entrepreneurs in Estonia themselves conceptualise the digital content they are commercialising, the nature of their business models, and entrepreneurial practice (business model). The authors take into account the applicable regulatory framework (protection of intellectual property, contract law, proprietary relations), and also analyse the contracting method, contractual remedies, and need for an objective neutrality in legal regulation.

Official bilingualism assumes the daily creation of parallel texts in bilingual areas. This is also the case in Istria County, where texts are usually written in Croatian and then translated into Italian. Given the official nature of texts and the context of the Italian language usage, it is extremely important to have a precise and uniform terminology, as well as well-developed language technologies, which would allow a faster and more accurate translation processes in bilingual institutions of Croatia. The aims of this paper are (1) to investigate the equal use of both Croatian and Italian as a minority language in the Istrian County through (1a) an analysis of the translated content on the official websites of the officially bilingual cities and municipalities and through (1b) an overview of current translation practices, and (2) to highlight the importance and necessity of the development of language technology through (2a) the presentation of trends in the development of language technology in similar bilingual and multilingual institutions, (2b) the presentation of the parallel corpus compiled from administrative texts of the Istrian County and (2c) the analysis of existing terminology conducted on the compiled corpus. The analysis of the availability of bilingual content on websites and the terminology extracted from the compiled parallel corpus indicates the need to develop and use translation tools and linguistic resources tailored to the Italian language as a minority language, in order to facilitate and accelerate the translation activity, thus allowing a more efficient use of Italian as a minority language. [ABSTRACT FROM AUTHOR]

Copyright of Letras (Lima) is the property of Letras, Facultad de Letras y Ciencias Humanas. Universidad Nacional Mayor De San Marcos and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Eesti Rakenduslingvistika Ühingu Aastaraamat is the property of Eesti Rakenduslingvistika Uhing (ERU) / Estonian Association for Applied Linguistics and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)