Language Corpora Research Articles

Questioni di filologia dantesca

This essay offers some linguistic remarks on food-related words in Dante’s Comedy, with particular attention to variants in the manuscript tradition. This preliminary analysis is related to the research conducted within the Project PRIN AtLiTeG (Atlante della lingua e dei testi della cultura gastronomica italiana dall’età medievale all’Unità), which aims to provide the first Corpus, Vocabulary and Atlas of ancient and modern Italian food language.
 

On creating a database of phrases for the language corpus

Ma’lumki, bugungi globallashgan davrda til korpuslari benazir o‘quv quroli, ko‘p funksiyali lingvistik elektron tizim hisoblanadi. Unda matnlar bazasi bilan birga leksikografik manbalarning ham mavjudligi ma’lumotlar manbai sifatida foydalanuvchilarga bisyor bilimga ega bo‘lish va so‘z boyliklarini oshirish imkoniyatlarini beradi. Shu maqsadda leksikografik baza tarkibida frazemalar bazasining bo‘lishi ahamiyatli sanaladi. Mazkur maqolada lingvistik korpuslarda iboralar bazasining mavjudligini ta’minlash sabablari va uning ahamiyati borasida so‘z yuritildi. Shuningdek, o‘zbek tilining ta’limiy korpusida iboralar bilan ishlash usullari ko‘rsatib berildi.

Axiological Foundations of Russian Statehood: “Truth” and “Justice” in the Domestic Ideological and Political Discourse

The concept of “truth” as a valuable spiritual and political category is constantly present in Russian spiritual and political writings from antiquity to the present day. But in the 18th century, there is an important transformation of this concept in the domestic ideological and political discourse: Under the influence of the Western European tradition in Russian socio-political writings of the state, conservative-educational and radical-educational trends, there is a gradual transition from traditional concepts, including the concept of “truth”, to such concepts as “justice” or “common use”. Thus, it acquires a significance that persists, including in modern political science literature. However, the concept of “truth” in its traditional meaning does not disappear at all from the socio-political discourse of the 18th century, but is used mainly in fiction, primarily in poetry, and in the writings of the traditionalist-Orthodox trend. A similar conclusion was made possible by an original methodological approach: the analysis of the works of the authors of the 18th century. It was carried out on the basis of the search database of the National Corpus of the Russian language, while in total more than 800 texts created in 1701-1800 were studied, in which the words “truth” and “justice” were used. In addition to the general scientific methodological apparatus, the main research methods were content analysis and political-textual analysis.

Kontekst historyczny wybranych kolokacji języka chorwackiego

The aim of the paper is to present the analysis of selected verb-noun collocations with the node ‘odluka’ – ‘decision’. The analysis was conducted on the basis of the 19th century newspaper articles published in Croatian newspapers from the period 1850–1899, collected in a digitalised format on the Stare hrvatske novine website. Historical forms are compared with the modern ones, excerpted from the HrWac Croatian language corpus. The second part of the 19th century is the time when Standard Croatian was being formed and the press was flourishing. This gave rise to the creation of the journalistic style which, alongside the administrative one, is believed to contain the greatest number of collocations, hence the choice of those two sources of data. The analysis of historical and modern forms has shown that there exist grammatical, orthographic, contextual, semantic and formal differences and these do not hinder the understanding of the text. The differences are negligent and stem from the evolution of the Croatian language, especially the shaping of the norm of the standard language.

BERT-5mC: an interpretable model for predicting 5-methylcytosine sites of DNA based on BERT

DNA 5-methylcytosine (5mC) is widely present in multicellular eukaryotes, which plays important roles in various developmental and physiological processes and a wide range of human diseases. Thus, it is essential to accurately detect the 5mC sites. Although current sequencing technologies can map genome-wide 5mC sites, these experimental methods are both costly and time-consuming. To achieve a fast and accurate prediction of 5mC sites, we propose a new computational approach, BERT-5mC. First, we pre-trained a domain-specific BERT (bidirectional encoder representations from transformers) model by using human promoter sequences as language corpus. BERT is a deep two-way language representation model based on Transformer. Second, we fine-tuned the domain-specific BERT model based on the 5mC training dataset to build the model. The cross-validation results show that our model achieves an AUROC of 0.966 which is higher than other state-of-the-art methods such as iPromoter-5mC, 5mC_Pred, and BiLSTM-5mC. Furthermore, our model was evaluated on the independent test set, which shows that our model achieves an AUROC of 0.966 that is also higher than other state-of-the-art methods. Moreover, we analyzed the attention weights generated by BERT to identify a number of nucleotide distributions that are closely associated with 5mC modifications. To facilitate the use of our model, we built a webserver which can be freely accessed at: http://5mc-pred.zhulab.org.cn.

Hate speech detection on Indonesian text using word embedding method-global vector

<p>Hate speech is defined as communication directed toward a specific individual or group that involves hatred or anger and a language with solid arguments leading to someone's opinion can cause social conflict. It has a lot of potential for individuals to communicate their thoughts on an online platform because the number of Internet users globally, including in Indonesia, is continually rising. This study aims to observe the impact of pre-trained global vector (GloVe) word embedding on accuracy in the classification of hate speech and non-hate speech. The use of pre-trained GloVe (Indonesian text) and single and multi-layer long short-term memory (LSTM) classifiers has performance that is resistant to overfitting compared to pre-trainable embedding for hatespeech detection. The accuracy value is 81.5% on a single layer and 80.9% on a double-layer LSTM. The following job is to provide pre-trained with formal and non-formal language corpus; pre-processing to overcome non-formal words is very challenging.</p>

20세기 전기 접속부사의 분포 양상과 담화 기능

The purpose of this paper is to study the distribution and functional characteristics of conjunctive adverbs appearing in spoken language corpus of the early 20th century. Through this, we presented the patterns in which conjunctive adverbs were used in spoken language and attempted to argue that the multi-layered connecting function of conjunctive adverbs was formed and continued during the early modern Korean language. Similar to the 21st century, conjunctive adverbs in the early 20th century had a distributional characteristic in which ‘그러-’ type conjunctive adverbs {그러면}, {그러나}, {그리고}, {그런데}, {그래서} appeared with high frequency. On the other hand, there were differences by period in the distribution of individual lists of conjunctive adverbs, and this was interpreted to be related to the formation period of individual conjunctive adverbs, register, and the expansion and contraction of their functions. Meanwhile conjunctive adverbs in the early 20th century also performed discourse functions in addition to general connecting functions. The discourse functions of conjunctive adverbs showed various aspects, including ‘utterance beginning’, ‘utterance finishing’, ‘utterance evading’, ‘filler’, ‘interruption’, ‘positive’, ‘utterance prompting’.

A Corpus Analysis of the Effects of Definiteness and Animacy on Word Order Variation

This article deals with the analysis of word order variation regarding subjects, direct objects, and non-direct object phrases called the “Target” in the corpus of languages of northwestern Iran, viz., Armenian, Mukri Kurdish, and Northeastern Kurdish (Indo-European), Jewish Northeastern Neo-Aramaic (Semitic), and Azeri Turkic (Turkic). The objective is to examine the effects of formal and semantic (in)definiteness in combination with animacy on Target word order variation to find out which one can be a triggering factor.

Bert Based Named Entity Recognition for the Albanian Language

In this paper, we explore the application of the Bidirectional Encoder Representations from Transformers (BERT) model for Named Entity Recognition (NER) in the Albanian language. Despite the success of BERT in various NLP tasks across numerous languages, Albanian remains under-represented. Our approach leverages a transfer learning strategy with multilingual BERT, fine-tuned on a newly created, comprehensive Albanian language corpus. The corpus includes texts from various domains, thereby facilitating the identification of a broad range of named entities. We report on the challenges faced during corpus creation, model fine-tuning, and testing phases, such as dealing with dialectal variations and lack of existing resources for the Albanian language. This research not only contributes to the advancement of NER systems for low-resource languages, but also provides a robust foundation for further advancements in Albanian language processing.
 
 Received: 21 September 2023 / Accepted: 29 October 2023 / Published: 23 November 2023

Language corpus and data driven learning (DDL) in language classrooms: A systematic review

This systematic review presents a comprehensive analysis of the literature on the use of corpora and data-driven learning (DDL) in language education. Corpus linguistics encompasses the use of electronic text collections for linguistic analysis, while DDL entails using corpora for pedagogical purposes in second/foreign language teaching. DDL allows language educators to move beyond traditional methods to enhance teaching practices and learning skills. An extensive database search identified 89 pertinent studies published between 1997 and 2022 that met the inclusion criteria. The selected studies focused on keywords such as "DDL," "corpus linguistics," and related phrases to identify relevant literature discussing DDL interventions in language classrooms. Only English, peer-reviewed texts with accessible PDFs were considered for inclusion. These studies described DDL implementation in classroom settings and the common pedagogical practices, difficulties, and limitations encountered. The findings suggest that DDL has significant potential as a pedagogical tool, but challenges exist that limit its positive impact on language learning. Tailored tasks, auxiliary guidance, supplemental support, and peer/group learning were identified as effective strategies for facilitating meaningful corpus engagement for lower-proficiency students.

Evolución semántica y lexicalización del somatismo español cerrar los ojos

The aim of this paper is to demonstrate the feasibility of applying the Invited Inferencing Theory of Se- mantic Change in order to establish the semantic evolution of idioms with literal correlates, in particular somatisms or kinegrams. The methodology is applied to analyze the testimonies (until 1700) of Spanish somatism cerrar los ojos in a diachronic corpus of Spanish language (CORDE), so as to find out what inferences generated contextual meanings that conditioned its lexicalization as idiom, semanticizing symbolic meanings related to death, sleep, knowledge, irrationality, submission or trust.

Transformer-Based Composite Language Models for Text Evaluation and Classification

Parallel natural language processing systems were previously successfully tested on the tasks of part-of-speech tagging and authorship attribution through mini-language modeling, for which they achieved significantly better results than independent methods in the cases of seven European languages. The aim of this paper is to present the advantages of using composite language models in the processing and evaluation of texts written in arbitrary highly inflective and morphology-rich natural language, particularly Serbian. A perplexity-based dataset, the main asset for the methodology assessment, was created using a series of generative pre-trained transformers trained on different representations of the Serbian language corpus and a set of sentences classified into three groups (expert translations, corrupted translations, and machine translations). The paper describes a comparative analysis of calculated perplexities in order to measure the classification capability of different models on two binary classification tasks. In the course of the experiment, we tested three standalone language models (baseline) and two composite language models (which are based on perplexities outputted by all three standalone models). The presented results single out a complex stacked classifier using a multitude of features extracted from perplexity vectors as the optimal architecture of composite language models for both tasks.

Corpus of Mandarin Child Language: a preliminary study on the acquisition of semantic content categories in Mandarin-speaking preschoolers.

In studying language acquisition in children, sizable research studies have been focusing on the investigation of form and lexical semantics. This study aims to establish a child language database annotated both syntactically with part of speech and semantically with semantic content category to supplement the study of child language acquisition in the semantic domain beyond lexical level. The Corpus of Mandarin Child Language (CMCL) that documented the production of different semantic content categories by Mandarin-speaking children was established. Naturalistic language samples of 82 native Mandarin-speaking children aged 25-60 months, divided into three age groups, were obtained. The corresponding semantic content categories coded in each utterance were tagged according to previous studies, in addition to the annotations of part of speech. MLU and lexical diversity were examined and the usage and acquisition of different semantic content categories were also analyzed. The results regarding syntactic complexity and lexical diversity replicated the typical language acquisition pattern from previous studies, which supported the validity of the data obtained in the CMCL. To investigate the trajectory of acquisition of various semantic content categories by age, a 90% acquisition criterion was used. Our findings regarding the acquisition order of semantic content category were basically in line with previous studies in general, with some minor differences. This acquisition order observed is largely explained by the cognitive and syntactic complexity associated with the semantic content category, with additional influence from language specific properties and cultural specific factors of Mandarin. In addition, with the tags in both part-of-speech and semantic content category, the CMCL potentially provides a platform for examining the form-content interface in early child language acquisition, which also implies significantly on the theoretical and clinical ground.

Precedence in Modern Spanish Journalism

The relevance of this article is due to the need to track the dynamics of the nature of the use of precedent units as an indicator of the level of linguistic competence in modern Spanish journalism. The purpose of the study is to identify the factors influencing the use of precedent units and trends in the dynamics of their content in modern Spanish journalism. The object of the study is precedent units, and the subject is their linguistic implementation in the modern Spanish press. The objectives of the study include the analysis of the representation of precedent units in modern media texts based on the study of stylistic characteristics, the definition of their content and the nature of their functioning in journalism. Any, even very minimal, shift in the value paradigm leads to a change in the content of the precedent unit, connections with its original content are lost and a complete deactualization of meaning may even occur. In the course of the study, methods of observation, definitional, quantitative, contextual, component and discourse analysis were used. The study was carried out based on Diccionario de dichos y frases hechas / "Dictionary of sayings and stable expressions" by Alberto Buitrago, the Corpus of the Spanish language of the XXI century (CORPES XXI) was used to track the frequency of use of precedent units. The article defines the tendency to change the content of precedent units in Spanish journalism. When updating the cultural content presented in the set of precedent units of Spanish linguoculture, the authors mainly use them as elements of headings that attract the reader's attention, while in the text of the article they often explain the meaning of this unit, which indicates the possible deactualization of this expression in the future.

How People With a Bipolar Disorder Diagnosis Talk About Personal Recovery in Peer Online Support Forums: Corpus Framework Analysis Using the POETIC Framework.

Personal recovery is of particular value in bipolar disorder, where symptoms often persist despite treatment. We previously defined the POETIC (Purpose and Meaning, Optimism and Hope, Empowerment, Tensions, Identity, Connectedness) framework for personal recovery in bipolar disorder. So far, personal recovery has only been studied in researcher-constructed environments (eg, interviews and focus groups). Support forum posts can serve as a complementary naturalistic data resource to understand the lived experience of personal recovery. This study aimed to answer the question "What can online support forum posts reveal about the experience of personal recovery in bipolar disorder in relation to the POETIC framework?" By integrating natural language processing, corpus linguistics, and health research methods, this study analyzed public, bipolar disorder support forum posts relevant to the lived experience of personal recovery. By comparing 4462 personal recovery-relevant posts by 1982 users to 25,197 posts not relevant to personal recovery, we identified 130 significantly overused key lemmas. Key lemmas were coded according to the POETIC framework. Personal recovery-related discussions primarily focused on 3 domains: "Purpose and meaning" (particularly reproductive decisions and work), "Connectedness" (romantic relationships and social support), and "Empowerment" (self-management and personal responsibility). This study confirmed the validity of the POETIC framework to capture personal recovery experiences shared on the web and highlighted new aspects beyond previous studies using interviews and focus groups. This study is the first to analyze naturalistic data on personal recovery in bipolar disorder. By indicating the key areas that people focus on in personal recovery when posting freely and the language they use, this study provides helpful starting points for formal and informal carers to understand the concerns of people diagnosed with a bipolar disorder and to consider how to best offer support.

Analyzing the Albanian Translations of Ernest Hemingway’s "The Old Man and the Sea" Ismail Kadare and Vedat Kokona: A Comparative Study

Translation, or more precisely translation studies, constitutes a very important topic nowadays and on which different theories continue to be thrown depending on the way of treatment and the importance given to the special aspects related to the translation process. Despite ongoing discussions, importance is attached to the fact that the translation should create an effect on the readers of the translated text as close as possible to that of the readers of the source text. This study aims to provide a detailed analysis based on the two versions translated into Albanian of the work The Old Man and the Sea by Hemingway, namely by Ismail Kadare in 1963 and by Vedat Kokona in 1997. The methodology of this study analyzes the features and characteristics of the aspects especially the two translated versions of the original work, emphasizing the similarities as well as the changes that the translation methodology has undergone over the years. By comparing the two different periods, this study brings to light the possible transformations that take place in the translation process, as well as in this way determine the course that the translation took until a few years ago. This study allows us to observe if those features continue to be part of the translation process even today. This research consists of two linguistic corpora, one in English and one in Albanian, which makes possible the realization of this work. Due to the nature of this work, the corpus in English is the work in the original language "The Old Man and the Sea" by Ernest Hemingway, as well as the Albanian language corpus, which constitutes the two versions of the translation of the work in the original language as well as scientific literature rich and contemporary in the service of elaborating this paper.
 
 Received: 11 September 2023 / Accepted: 20 October 2023 / Published: 5 November 2023

Par dažu bezpriedēkļa un priedēkļa verbu lietojumu ar adverbiem un prievārdu pār

In Latvian language, prefix and prefix-free verb constructions with prepositions and adverbs are used in parallel, where the prefix, adverb and preposition are close in terms of the meaning. The aim of the research is to study the connection of prefix-free and prefix verbs with prepositions and adverbs, to study the use of these constructions, reciprocal attitudes and semantics. The novelty of the research is indicated by the fact that extensive systematic studies have not been conducted on the connection between prefixed and prefix-free verb and preposition and/or adverb, their semantics, as well as the possible motivation for doubling the preposition and adverb. The article examines constructions with prepositions uz, no, caur, pār, expressing spatial direction attitudes and which, in modern Latvian, have prefixes corresponding in the meaning – uz-, no-, pār-, as well as adverbs: virsū, ārā, cauri, pāri, nost. The source of the examples analysed in the study is written language on the Internet – chatting, i.e. microblogging site „Twitter”, news publications, comments on the Internet, „Web Corpus of Latvian Language” (Tīmeklis2020) and other sources where coverage of spoken language is available. It has been observed that in cases when a prefix, adverb and preposition of close meaning are used in the same sentence, it is easier to visualize what is being said as if it is being „drawn” before the eyes, and such structures make the text more imaginative. The meaning of a verb plays an important role in what adverb or preposition it binds to itself. In the research, a tendency was observed that most of the time, in constructions of prefixed verb and prefix-free verb, preposition and adverb with verbs of motion of close meaning and of literal sense, are used.

Event Knowledge in Large Language Models: The Gap Between the Impossible and the Unlikely.

Word co-occurrence patterns in language corpora contain a surprising amount of conceptual knowledge. Large language models (LLMs), trained to predict words in context, leverage these patterns to achieve impressive performance on diverse semantic tasks requiring world knowledge. An important but understudied question about LLMs' semantic abilities is whether they acquire generalized knowledge of common events. Here, we test whether five pretrained LLMs (from 2018's BERT to 2023's MPT) assign a higher likelihood to plausible descriptions of agent-patient interactions than to minimally different implausible versions of the same event. Using three curated sets of minimal sentence pairs (total n = 1215), we found that pretrained LLMs possess substantial event knowledge, outperforming other distributional language models. In particular, they almost always assign a higher likelihood to possible versus impossible events (The teacher bought the laptop vs. The laptop bought the teacher). However, LLMs show less consistent preferences for likely versus unlikely events (The nanny tutored the boy vs. The boy tutored the nanny). In follow-up analyses, we show that (i) LLM scores are driven by both plausibility and surface-level sentence features, (ii) LLM scores generalize well across syntactic variants (active vs. passive constructions) but less well across semantic variants (synonymous sentences), (iii) some LLM errors mirror human judgment ambiguity, and (iv) sentence plausibility serves as an organizing dimension in internal LLM representations. Overall, our results show that important aspects of event knowledge naturally emerge from distributional linguistic patterns, but also highlight a gap between representations of possible/impossible and likely/unlikely events.

Stem overgeneralizations in the acquisition of Croatian verbal morphology: Evidence from parental questionnaires

Studies on verbal overgeneralization often focus on languages with low morphological complexity. The Croatian conjugational system exhibits varying degrees of complexity, and this complexity is not primarily based on the number of inflectional morphemes, but on an elaborate system of stem changes. During early language development, children face the difficult task of acquiring this system, using overgeneralized forms to overcome its complexity. To date, studies have used a corpus-based method to retrieve overgeneralizations in child language, which has had limited success in capturing this phenomenon. The aim of this study was to investigate the production of overgeneralized verb forms in Croatian monolingual children aged 2;6 to 5;11 using a questionnaire in which parents report overgeneralizations used by their children. We tested the relationship between the production of overgeneralized forms and features of the input language (token frequency and class size). We hypothesized that the rate of overgeneralizations will depend on input language features, i.e. a higher rate of overgeneralizations for infrequent verbs and for verbs with smaller class size. The items selected for the questionnaire are the verbs with stem change used by parents in the longitudinal Croatian corpus of child language. Parents report overgeneralized forms in all verb classes, and verb frequency and class size negatively correlate with the proportion of overgeneralizations. Our results show that children gradually abstract morphological systems in a way that is highly sensitive to the properties of the input.

Studying negative evidence in Finnish language corpora

This study explores the relationship between lower-than-expected frequencies of word forms and inherent gaps in Finnish inflectional paradigms. The research aims to determine whether it is possible to predict paradigmatic gaps from lower-than-expected frequencies of word forms. We examined Finnish nouns inflected in a marginal case (the instructive) and hypothesized that some of these nouns may potentially have gaps in their inflectional paradigms. However, we found that such gaps are contingent and do not cause uncertainty when filled. We find that the correlation between inherent gaps and lower frequencies is one-directional: predicting inherent gaps from lower-than-expected frequencies is problematic. The results suggest that any paradigmatic gap suggested by corpus frequency is more likely to be contingent than inherent, and that the less semantic need there is for a particular word form, the more likely it will be unattested even in a large corpus. The research highlights the importance of considering semantic profiles when analyzing the grammaticality of word forms and suggests that statistical tests like Fisher’s exact are not necessarily the right approach to tackle the problem of negative evidence in corpus studies.