Hasil untuk "Language. Linguistic theory. Comparative grammar"

Olga V. Sergeeva, Marina R. Zheltukhina, Elena B. Ponomarenko

In the 21st century, the interdisciplinary research interest is increasingly aroused by the linguistic problem of realizing educational discourse, especially in the media space. Information ecology and cyber security make it possible to form the skill of making adequate educational, and managerial decisions in the field of education under the conditions of saturation with information, information noise in the digital environment. The purpose of the study is to identify the genre features of the educational media discourse in the context of information ecology and cyber security. A functional-genre analysis of the educational media discourse in the digital media space is carried out in the study. By applying a set of methods (descriptive method, content analysis, discursive analysis, linguosemiotic analysis, linguopragmatic analysis, functional-genre analysis, interpretive analysis), the genre media reflection of the regulation of society’s activities to achieve hygienic safety goals related to information is studied, which constitutes the scientific novelty of the study. The main genres of the educational media discourse that are significant for the development of genre theory have been identified, those are: analytical, popularizing, explanatory, didactic, regulating, recommendation, discussion, multimedia, case study genres. Their analyses allow conclude that an ecosystem comfortable for training students is formed due to compliance with the requirements of information ecology, information and Internet hygiene, ethics, cyber security, which are among the preventive trends and protective measures in the digital environment. The analysis of factual material emphasizes the importance of ensuring the safety of students as one of the key tasks of the modern educational process, considering the active influence of the media environment. It is established that media articles inform an addressee about the activities of preventive medicine and state sanitary and epidemiological services that are developing norms that reflect the safe organization of the work and educational process using information tools in the digital media space. Documents presented in various media genres determine the norms of lighting at different times of the day, the norms of noise parameters and work with electronic teaching aids and other acceptable conditions to ensure high-quality work without harm to health. The identified genre features of the educational media discourse in the context of information ecology and cyber security clearly demonstrate that the informational ecology is a promising direction for the study and development of the media discourse, incl. educational media discourse, based on the material of various linguistic cultures.

Héctor Leví Caballero Artigas

Cong-Thanh Do, Rama Doddipatla, Kate Knill

Chain-of-Thought (CoT) prompting is a widely used method to improve the reasoning capability of Large Language Models (LLMs). More recently, CoT has been leveraged in Knowledge Distillation (KD) to transfer reasoning capability from a larger LLM to a smaller one. This paper examines the role of CoT in distilling the reasoning capability from larger LLMs to smaller LLMs using white-box KD, analysing its effectiveness in improving the performance of the distilled models for various natural language reasoning and understanding tasks. We conduct white-box KD experiments using LLMs from the Qwen and Llama2 families, employing CoT data from the CoT-Collection dataset. The distilled models are then evaluated on natural language reasoning and understanding tasks from the BIG-Bench-Hard (BBH) benchmark, which presents complex challenges for smaller LLMs. Experimental results demonstrate the role of CoT in improving white-box KD effectiveness, enabling the distilled models to achieve better average performance in natural language reasoning and understanding tasks from BBH.

Aadi Palnitkar, Arjun Suresh, Rishi Rajesh et al.

Increasingly, more and more people are turning to large language models (LLMs) for healthcare advice and consultation, making it important to gauge the efficacy and accuracy of the responses of LLMs to such queries. While there are pre-existing medical benchmarks literature which seeks to accomplish this very task, these benchmarks are almost universally in English, which has led to a notable gap in existing literature pertaining to multilingual LLM evaluation. Within this work, we seek to aid in addressing this gap with Cancer-Myth-Indic, an Indic language benchmark built by translating a 500-item subset of Cancer-Myth, sampled evenly across its original categories, into five under-served but widely used languages from the subcontinent (500 per language; 2,500 translated items total). Native-speaker translators followed a style guide for preserving implicit presuppositions in translation; items feature false presuppositions relating to cancer. We evaluate several popular LLMs under this presupposition stress.

Ramatu Oiza Abdulsalam, Segun Aroyehun

Recent work has explored the use of large language models (LLMs) to generate tutoring responses in mathematics, yet it remains unclear how closely their instructional behavior aligns with expert human practice. We analyze a dataset of math remediation dialogues in which expert tutors, novice tutors, and seven LLMs of varying sizes, comprising both open-weight and commercial models, respond to the same student errors. We examine instructional strategies and linguistic characteristics of tutoring responses, including uptake (restating and revoicing), pressing for accuracy and reasoning, lexical diversity, readability, politeness, and agency. We find that expert tutors produce higher-quality responses than novices, and that larger LLMs generally receive higher pedagogical quality ratings than smaller models, approaching expert performance on average. However, LLMs exhibit systematic differences in their instructional profiles: they underuse discursive strategies characteristic of expert tutors while generating longer, more lexically diverse, and more polite responses. Regression analyses show that pressing for accuracy and reasoning, restating and revoicing, and lexical diversity, are positively associated with perceived pedagogical quality, whereas higher levels of agentic and polite language are negatively associated. These findings highlight the importance of analyzing instructional strategies and linguistic characteristics when evaluating tutoring responses across human tutors and intelligent tutoring systems.

Ioana Buhnila, Aman Sinha, Mathieu Constant

Large Language Models (LLMs) have proven efficient in giving definition-type answers to user input queries. While for humans giving various types of answers, such as examples and paraphrases, is an easy task, LLMs struggle to provide correct answers for other than definition-type queries. In this study, we evaluated this drop in performance using TrackList, a fine-grained linguistic and statistical analysis pipeline to investigate the impact of the pre-training data on LLMs answers to diverse linguistic queries. We also introduce RefoMed-EN, an English dataset consisting of 6170 human-annotated medical terms alongside their corresponding definitions, denominations, exemplifications, explanations, or paraphrases. We studied whether the high frequency of a concept (head) or low frequency (tail) impacts the language model's performance. We evaluated the quality of the LLM's output using syntactic and semantic similarity metrics, statistical correlations and embeddings. Results showed that the LLM's task performance for definition type questions is the highest, while for the exemplification type it is the lowest. Additionally, we showed that for definition-type questions, large language models are prone to paraphrase more on popular and frequent knowledge and less on tail and technical knowledge, especially in the expert texts.

Reza Nazemian, Ahmad Mortazavi

Abstract The significance of speakers' objectives in political writing renders the techniques utilized by authors especially important. The translation method is essential because of the strong relationship between persuasive tools and factors such as the author's or narrator's reputation, their role in shaping ideas, and the vision created in the audience's mind to attain certain objectives. Ethos is a rhetorical strategy that facilitates text comprehension and ensures precise translation. This study employs a descriptive-analytical method, focusing on three primary axes: the text creator, the audience, and the complex influence of ethos on text production and reproduction. It is vital to employ a suitable technique in replicating the text to guarantee that the primary point is communicated clearly and articulately. This is due to the significant emphasis on the author's position and standing, the representation of their cultural identity, the translator's methodologies, and the satisfaction of the readers' anticipations of the communicated message. This analysis focuses on a work by Alaa Al-Aswany, a modern Egyptian writer, titled Our Advice to the Butcher. Findings show that paying close attention to the author's subtleties and nuances in word choice, such as the choice of conjunctions, as well as finding equivalents for vocabulary and grammatical structures, appealing to the reader's emotions, and using clear and effective language during the translation process are all very important for a successful outcome.IntroductionIn rhetoric, the speaker or writer aims to create an image that captivates the audience and persuades them, fostering a reciprocal communication exchange between the writer and the listener. We discover a collection of strategies and techniques when we analyze the works of orators and novelists that emphasize raising awareness and addressing public opinion while meticulously using the expressive potential of language. These methods and techniques are utilized to facilitate effective communication with the audience and to persuade them to embrace the intended concepts.Ethos, as a language strategy, has a historical foundation, and analyzing this idea might facilitate a clear comprehension of certain texts. It is very proficient in generating a valuable and suitable translation of the material while effectively expressing the nuances pertinent to the audience. Without knowledge or attention to this topic, the primary message of the text may not be adequately conveyed to the audience. By emphasizing pivotal words and terms that activate the audience's psychological elements, the translator can facilitate their gradual acceptance of the knowledge and the adoption of the intended actions. This method enhances the audience's viewpoint, the reinterpretation of the text, and their resolve to embrace or dismiss the concepts presented.Statement of the ProblemThe point of this study is to look into ethos as a persuasive tool and how it works in translation, as well as how it shows up in the steps of figuring out equivalence and choosing preferred language and expressions. The narrative "Our Advice to the Butcher" (originally "Nasihatuna ila al-Jazzar"), extracted from the anthology "Do We Deserve Democracy?" (originally "Hal Nastaḥiq al-Dimuqratiyah?"), functions as the case study for this analysis. Employing the ethos method, we examine the linguistic features and translation techniques of the Arabic text into Persian.  Literature ReviewResearch in linguistics is a longstanding and comprehensive discipline. The same applies to text translation and its diverse methodologies, which have garnered the interest of numerous contemporary professors, students, and researchers. The domain of rhetorical techniques has been the subject of numerous books and studies. The primary source in this domain is Aristotle's (1979) Rhetoric. In this book, Aristotle characterizes ethos as a picture that emerges subsequent to discourse within the context of rhetoric. The rhetorical triangle, a notion established by the Greek philosopher Aristotle, encompasses three essential components of persuasive arguments: logos, ethos, and pathos. Numerous essays and theses have been written analyzing the works of the renowned contemporary Egyptian novelist Alaa Al-Aswany; however, none have examined the translation of his works through the lens of the ethos method. Consequently, the present study represents the inaugural endeavor in this domain. No article has yet been published regarding the application of the ethos technique in translation, and, to the researchers' knowledge, it remains unexamined.MethodologyThis study employs a descriptive-analytical methodology grounded on structuralism and incorporates the theories of substitution and collocation. It analyzes different forms of ethos as a potent discursive strategy in communicative processes that influence the subconscious to perform speech acts and achieve pragmatic goals. Furthermore, it examines the function of ethos in the text translation process and analyzes diverse methodologies.ConclusionThe ethos strategy highlights three primary dimensions: the text creator, the audience, and the diverse effects on text generation and replication. It is vital to employ a suitable technique in replicating the text to guarantee that the primary point is communicated clearly and articulately. This is due to ethos placing significant emphasis on the author's position and standing, the translation of their cultural identity, and the satisfaction of readers' expectations for the imparted message.The element of symbolism is notably evident in the collection Do We Deserve Democracy? by modern Egyptian novelist Alaa Al-Aswany. The author contends that democracy in Egypt operates solely as a façade and lacks any substance.The ethos method relies on convincing the reader. Therefore, to maintain the original author's stance and cultural identity, the translation must align the grammatical structures of the source text. The translated text must be believable and persuasive to the audience.The translated material must employ clear and compelling language, necessitating alterations in verb forms and tenses. Linguistic tools are essential for conveying the meaning, subtleties, and complexity embedded in the original text by the author. Every language has distinct powers and potentials, and the translator must comprehend the linguistic tools of Arabic to identify parallels in Persian. This guarantees, firstly, that the tools and grammatical structures of the source language do not encroach upon the target language, and secondly, that Persian readers experience a sense of familiarity and intimacy with the translated text, rendering it credible as though they are reading a text in their own language.Idiomatic expressions, phrases, and the names of locations and cultural situations must be translated to maintain the authenticity of the original author and to accurately represent their cultural identity in the target language text.

Michela Venditti

Il saggio indaga le relazione tra il modernismo russo, in particolare l'opera di V. Brjusov, e la filosofia di Nikolaj Fedorov e del suo divulgatore K. Ciolkovskij. L'analisi dei testi, in prosa e in versi, del poeta simbolista è preceduta da un breve excursus sulle origini della fantascienza russa. Brjusov è uno dei primi teorici russi della fantascienza, che definisce come genere nell'articolo inedito "I confini della fantastika". Nel periodo post-rivoluzionario il cosmismo russo si manifesta soprattutto nella idea della conquista dello spazio cosmico, alla base della fantascienza sovietica. Brjusov, uno dei pochi ad aderire alla rivoluzione d'Ottobre, segue nella sua opera questa evoluzione.

Indah Mauludina, Ayesa Ayesa, Tri Wahyu Retno Ningsih

<p>This study aims to investigate white wine in various solemnities performed in Tridharma temples. The data were collected by observation, interview, and documentation. To obtain data, the informants were interviewed. The informants were one <em>biokong</em> and two worshippers of Tridharma temples. The total number of informants was six. Two Tridharma temples were chosen as the locations for conducting this research. It was Toa Se Bio Temple. As it is the oldest Tridharma temple in West Java, the writer chose it as the location of the study. The theory used in analyzing the data were Xiao’s theory (1995), a textbook entitled <em>The</em> <em>Worshipper Leaders</em> (2015), and a book entitled <em>A Brief History of Sian Djin Ku Poh Temple</em> (2019). The results revealed that white wine has been used in worship. It is used in worship toward ancestors and sinbengs. There were two kinds of the usage of white wine. The use of white wine in prayer activities is still valid, especially for prayer activities and festivals. As for prayer activities that still use white wine, namely prayers to worship sinbeng, <em>Cap Go Meh</em> night activities, and <em>Sejit</em>. The use of white wine in prayer activities is a tradition that has existed for generations. It is also used in tangsin ceremony, Cap Go Meh Eve, and sejit ceremony.</p><p class="abstrak"> </p><p class="abstrak"><strong>Received: 29 August 2023 </strong></p><p class="abstrak"><strong>Accepted: 06 January 2024 </strong></p><p class="abstrak"><strong>Published: 24 April 2024</strong></p>

Akash Ghosh, Arkadeep Acharya, Sriparna Saha et al.

The advent of Large Language Models (LLMs) has significantly reshaped the trajectory of the AI revolution. Nevertheless, these LLMs exhibit a notable limitation, as they are primarily adept at processing textual information. To address this constraint, researchers have endeavored to integrate visual capabilities with LLMs, resulting in the emergence of Vision-Language Models (VLMs). These advanced models are instrumental in tackling more intricate tasks such as image captioning and visual question answering. In our comprehensive survey paper, we delve into the key advancements within the realm of VLMs. Our classification organizes VLMs into three distinct categories: models dedicated to vision-language understanding, models that process multimodal inputs to generate unimodal (textual) outputs and models that both accept and produce multimodal inputs and outputs.This classification is based on their respective capabilities and functionalities in processing and generating various modalities of data.We meticulously dissect each model, offering an extensive analysis of its foundational architecture, training data sources, as well as its strengths and limitations wherever possible, providing readers with a comprehensive understanding of its essential components. We also analyzed the performance of VLMs in various benchmark datasets. By doing so, we aim to offer a nuanced understanding of the diverse landscape of VLMs. Additionally, we underscore potential avenues for future research in this dynamic domain, anticipating further breakthroughs and advancements.

Yu Zhao, Hao Fei, Shengqiong Wu et al.

Grammar Induction could benefit from rich heterogeneous signals, such as text, vision, and acoustics. In the process, features from distinct modalities essentially serve complementary roles to each other. With such intuition, this work introduces a novel \emph{unsupervised visual-audio-text grammar induction} task (named \textbf{VAT-GI}), to induce the constituent grammar trees from parallel images, text, and speech inputs. Inspired by the fact that language grammar natively exists beyond the texts, we argue that the text has not to be the predominant modality in grammar induction. Thus we further introduce a \emph{textless} setting of VAT-GI, wherein the task solely relies on visual and auditory inputs. To approach the task, we propose a visual-audio-text inside-outside recursive autoencoder (\textbf{VaTiora}) framework, which leverages rich modal-specific and complementary features for effective grammar parsing. Besides, a more challenging benchmark data is constructed to assess the generalization ability of VAT-GI system. Experiments on two benchmark datasets demonstrate that our proposed VaTiora system is more effective in incorporating the various multimodal signals, and also presents new state-of-the-art performance of VAT-GI.

Michael Duan, Anshuman Suri, Niloofar Mireshghallah et al.

Membership inference attacks (MIAs) attempt to predict whether a particular datapoint is a member of a target model's training data. Despite extensive research on traditional machine learning models, there has been limited work studying MIA on the pre-training data of large language models (LLMs). We perform a large-scale evaluation of MIAs over a suite of language models (LMs) trained on the Pile, ranging from 160M to 12B parameters. We find that MIAs barely outperform random guessing for most settings across varying LLM sizes and domains. Our further analyses reveal that this poor performance can be attributed to (1) the combination of a large dataset and few training iterations, and (2) an inherently fuzzy boundary between members and non-members. We identify specific settings where LLMs have been shown to be vulnerable to membership inference and show that the apparent success in such settings can be attributed to a distribution shift, such as when members and non-members are drawn from the seemingly identical domain but with different temporal ranges. We release our code and data as a unified benchmark package that includes all existing MIAs, supporting future work.

Oline Ranum, Gomer Otterspeer, Jari I. Andersen et al.

In this work, we present an efficient approach for capturing sign language in 3D, introduce the 3D-LEX v1.0 dataset, and detail a method for semi-automatic annotation of phonetic properties. Our procedure integrates three motion capture techniques encompassing high-resolution 3D poses, 3D handshapes, and depth-aware facial features, and attains an average sampling rate of one sign every 10 seconds. This includes the time for presenting a sign example, performing and recording the sign, and archiving the capture. The 3D-LEX dataset includes 1,000 signs from American Sign Language and an additional 1,000 signs from the Sign Language of the Netherlands. We showcase the dataset utility by presenting a simple method for generating handshape annotations directly from 3D-LEX. We produce handshape labels for 1,000 signs from American Sign Language and evaluate the labels in a sign recognition task. The labels enhance gloss recognition accuracy by 5% over using no handshape annotations, and by 1% over expert annotations. Our motion capture data supports in-depth analysis of sign features and facilitates the generation of 2D projections from any viewpoint. The 3D-LEX collection has been aligned with existing sign language benchmarks and linguistic resources, to support studies in 3D-aware sign language processing.

Iuliia Zaitova, I. Stenger, T. Avgustinova

Microsyntactic units have been defined as language-specific transitional entities between lexicon and grammar, whose idiomatic properties are closely tied to syntax. These units are typically described based on individual constructions, making it difficult to understand them comprehensively as a class. This study proposes a novel approach to detect microsyntactic units using Word Embedding Models (WEMs) trained on six Slavic languages, namely Belarusian, Bulgarian, Czech, Polish, Russian, and Ukrainian, and evaluates how well these models capture the nuances of syntactic non-compositionality. To evaluate the models, we develop a cross-lingual inventory of microsyntactic units using the lists of microsyntantic units available at the Russian National Corpus. Our results demonstrate the effectiveness of WEMs in capturing microsyntactic units across all six Slavic languages under analysis. Additionally, we find that WEMs tailored for syntax-based tasks consistently outperform other WEMs at the task. Our findings contribute to the theory of microsyntax by providing insights into the detection of microsyntactic units and their cross-linguistic properties.

Liina Pylkkänen

Maksud Sharipov, Elmurod Kuriyozov, Ollabergan Yuldashev et al.

This research paper presents a part-of-speech (POS) annotated dataset and tagger tool for the low-resource Uzbek language. The dataset includes 12 tags, which were used to develop a rule-based POS-tagger tool. The corpus text used in the annotation process was made sure to be balanced over 20 different fields in order to ensure its representativeness. Uzbek being an agglutinative language so the most of the words in an Uzbek sentence are formed by adding suffixes. This nature of it makes the POS-tagging task difficult to find the stems of words and the right part-of-speech they belong to. The methodology proposed in this research is the stemming of the words with an affix/suffix stripping approach including database of the stem forms of the words in the Uzbek language. The tagger tool was tested on the annotated dataset and showed high accuracy in identifying and tagging parts of speech in Uzbek text. This newly presented dataset and tagger tool can be used for a variety of natural language processing tasks such as language modeling, machine translation, and text-to-speech synthesis. The presented dataset is the first of its kind to be made publicly available for Uzbek, and the POS-tagger tool created can also be used as a pivot to use as a base for other closely-related Turkic languages.

Marcus Maia

The present study reports and discusses the use of eye tracking qualitative data (dynamic gaze plots and heatmaps) in reading workshops in a middle school and in Generative Syntax and Sentence Processing courses at the undergraduate and graduate level. Both endeavors take the sentential level as the proper object to be metacognitively explored in language education in order to develop innate science forming capacity and knowledge of language. In both projects non-discrepant qualitative eye tracking data collected and quantitatively analyzed in psycholinguistic studies carried out in Lapex (Experimental Psycholinguistics Laboratory of the Federal University of Rio de Janeiro) were displayed to students as a point of departure, triggering discussions. Active, problem-solving based methodologies were employed with the objective of stimulating student participation. The article also discusses the importance of developing full literacy, epistemic vigilance and intellectual self-defense in an infodemic world.

Zhihong Chen, Yuhao Du, Jinpeng Hu et al.

Medical vision-and-language pre-training provides a feasible solution to extract effective vision-and-language representations from medical images and texts. However, few studies have been dedicated to this field to facilitate medical vision-and-language understanding. In this paper, we propose a self-supervised learning paradigm with multi-modal masked autoencoders (M$^3$AE), which learn cross-modal domain knowledge by reconstructing missing pixels and tokens from randomly masked images and texts. There are three key designs to make this simple approach work. First, considering the different information densities of vision and language, we adopt different masking ratios for the input image and text, where a considerably larger masking ratio is used for images. Second, we use visual and textual features from different layers to perform the reconstruction to deal with different levels of abstraction in visual and language. Third, we develop different designs for vision and language decoders (i.e., a Transformer for vision and a multi-layer perceptron for language). To perform a comprehensive evaluation and facilitate further research, we construct a medical vision-and-language benchmark including three tasks. Experimental results demonstrate the effectiveness of our approach, where state-of-the-art results are achieved on all downstream tasks. Besides, we conduct further analysis to better verify the effectiveness of different components of our approach and various settings of pre-training. The source code is available at~\url{https://github.com/zhjohnchan/M3AE}.

Edwin Zhang, Yujie Lu, Shinda Huang et al.

Training generalist agents is difficult across several axes, requiring us to deal with high-dimensional inputs (space), long horizons (time), and generalization to novel tasks. Recent advances with architectures have allowed for improved scaling along one or two of these axes, but are still computationally prohibitive to use. In this paper, we propose to address all three axes by leveraging \textbf{L}anguage to \textbf{C}ontrol \textbf{D}iffusion models as a hierarchical planner conditioned on language (LCD). We effectively and efficiently scale diffusion models for planning in extended temporal, state, and task dimensions to tackle long horizon control problems conditioned on natural language instructions, as a step towards generalist agents. Comparing LCD with other state-of-the-art models on the CALVIN language robotics benchmark finds that LCD outperforms other SOTA methods in multi-task success rates, whilst improving inference speed over other comparable diffusion models by 3.3x~15x. We show that LCD can successfully leverage the unique strength of diffusion models to produce coherent long range plans while addressing their weakness in generating low-level details and control.

J. Spencer

In Telling Animals, Jasmine Spencer offers a comparative yet personal approach to Dene/Athabaskan stories, both Northern and Southern. It examines the animating effects of animal stories, the transformative power of animacies in Dene stories, and the effects of narrative revitalization through animal grammar. It takes as its first premise the teachings of many Elders, who have shared that the stories are alive. Jasmine Spencer's comparative approach combines literary, linguistic, anthropological, and philosophical theories and methods using a deictic framework for closely reading the stories in both their Dene languages and in English translation. The narrative epistemologies enacted by Dene stories counterbalance many of the ethical problems inherent within Euro-Western approaches to ontology and experience. These stories revive those who listen and read, offering hope.