Sign language translation systems typically require English as an intermediary language, creating barriers for non-English speakers in the global deaf community. We present Canonical Semantic Form (CSF), a language-agnostic semantic representation framework that enables direct translation from any source language to sign language without English mediation. CSF decomposes utterances into nine universal semantic slots: event, intent, time, condition, agent, object, location, purpose, and modifier. A key contribution is our comprehensive condition taxonomy comprising 35 condition types across eight semantic categories, enabling nuanced representation of conditional expressions common in everyday communication. We train a lightweight transformer-based extractor (0.74 MB) that achieves 99.03% average slot extraction accuracy across four typologically diverse languages: English, Vietnamese, Japanese, and French. The model demonstrates particularly strong performance on condition classification (99.4% accuracy) despite the 35-class complexity. With inference latency of 3.02ms on CPU, our approach enables real-time sign language generation in browser-based applications. We release our code, trained models, and multilingual dataset to support further research in accessible sign language technology.
This paper evaluates the performance of transformer-based language models on split-ergative case alignment in Georgian, a particularly rare system for assigning grammatical cases to mark argument roles. We focus on subject and object marking determined through various permutations of nominative, ergative, and dative noun forms. A treebank-based approach for the generation of minimal pairs using the Grew query language is implemented. We create a dataset of 370 syntactic tests made up of seven tasks containing 50-70 samples each, where three noun forms are tested in any given sample. Five encoder- and two decoder-only models are evaluated with word- and/or sentence-level accuracy metrics. Regardless of the specific syntactic makeup, models performed worst in assigning the ergative case correctly and strongest in assigning the nominative case correctly. Performance correlated with the overall frequency distribution of the three forms (NOM > DAT > ERG). Though data scarcity is a known issue for low-resource languages, we show that the highly specific role of the ergative along with a lack of available training data likely contributes to poor performance on this case. The dataset is made publicly available and the methodology provides an interesting avenue for future syntactic evaluations of languages where benchmarks are limited.
Charlotte Dumont, Emma Peri, Arnaud Destrebecqz
et al.
Background and aims Language development in autism varies widely, from fluently verbal to minimally verbal individuals, with socio-communicative difficulties often cited as key explanatory factors. Statistical learning (SL)—the ability to detect regularities in language—has also emerged as a potential contributor to language acquisition in autism. However, SL research in autism has predominantly focused on verbally fluent individuals, leaving non- and minimally verbal populations underexplored. This study aimed to examine the predictive roles of joint attention and statistical learning, specifically nonadjacent dependency learning, on expressive vocabulary and morphosyntactic outcomes in autistic children. Methods Participants included 40 autistic children aged 5–8 years with diverse linguistic profiles, ranging from verbally fluent to minimally verbal, and 40 non-autistic children. Joint attention was assessed during a semi-structured play protocol, which also provided naturalistic language samples for analysis. Measures of expressive vocabulary and morphosyntax were derived from the number of different words and verb flexions produced, respectively. Sensitivity to nonadjacent dependencies was evaluated through an artificial language learning task. Results Neither joint attention nor sensitivity to nonadjacent dependencies predicted expressive vocabulary or morphosyntactic skills in autistic children. Response to joint attention scores were significantly lower in autistic children than in non-autistic children but higher than in previous research. This may be due to the less structured and, therefore, more ecologically valid context in which joint attention was assessed (free play), in conjunction with age and maturation factors. Regarding the SL task, both autistic and non-autistic children demonstrated sensitivity to nonadjacent dependencies. Most interestingly perhaps, only 15 autistic children completed the SL task, with non-verbal cognitive abilities significantly predicting task completion. Conclusions and implications This study highlights the complexity of investigating the role of statistical learning in language development in autism. It underscores the limitations of behavioral SL paradigms for minimally verbal children. Future research should prioritize developing more ecologically valid and accessible paradigms to accurately assess statistical learning in minimally verbal children, thereby clarifying the role SL may play in language acquisition in autism.
Special aspects of education, Neurosciences. Biological psychiatry. Neuropsychiatry
In recent years, the breakthrough of Large Language Models (LLMs) offers new ideas for achieving universal methods on graph data. The common practice of converting graphs into natural language for LLMs, which refers to graph flattening, exhibits good generalizability and interpretability. However, the poor organization of the textual format results in poor performance in long-distance scenario understanding. Inspired by human cognitive reasoning habits, we propose a novel method for graph flattening to fit LLMs, termed as End-to-End DAG-Path prompting (EEDP). Experiments on real-world datasets show that EEDP enhances the reasoning performance of LLMs in long-distance scenarios while maintaining excellent performance in short-distance scenarios, demonstrating good robustness in the face of distance variations.
The rapid growth of Large Language Models (LLMs) has put forward the study of biases as a crucial field. It is important to assess the influence of different types of biases embedded in LLMs to ensure fair use in sensitive fields. Although there have been extensive works on bias assessment in English, such efforts are rare and scarce for a major language like Bangla. In this work, we examine two types of social biases in LLM generated outputs for Bangla language. Our main contributions in this work are: (1) bias studies on two different social biases for Bangla, (2) a curated dataset for bias measurement benchmarking and (3) testing two different probing techniques for bias detection in the context of Bangla. This is the first work of such kind involving bias assessment of LLMs for Bangla to the best of our knowledge. All our code and resources are publicly available for the progress of bias related research in Bangla NLP.
This chapter focuses on gender-related errors in machine translation (MT) in the context of low-resource languages. We begin by explaining what low-resource languages are, examining the inseparable social and computational factors that create such linguistic hierarchies. We demonstrate through a case study of our mother tongue Bengali, a global language spoken by almost 300 million people but still classified as low-resource, how gender is assumed and inferred in translations to and from the high(est)-resource English when no such information is provided in source texts. We discuss the postcolonial and societal impacts of such errors leading to linguistic erasure and representational harms, and conclude by discussing potential solutions towards uplifting languages by providing them more agency in MT conversations.
Specific Language Impairment (SLI) is a disorder that affects communication and can affect both comprehension and expression. This study focuses on effectively detecting SLI in children using transcripts of spontaneous narratives from 1063 interviews. A three-stage cascading pipeline was proposed f. In the first stage, feature extraction and dimensionality reduction of the data are performed using the Random Forest (RF) and Spearman correlation methods. In the second stage, the most predictive variables from the first stage are estimated using logistic regression, which is used in the last stage to detect SLI in children from transcripts of spontaneous narratives using a nearest neighbor classifier. The results revealed an accuracy of 97.13% in identifying SLI, highlighting aspects such as the length of the responses, the quality of their utterances, and the complexity of the language. This new approach, framed in natural language processing, offers significant benefits to the field of SLI detection by avoiding complex subjective variables and focusing on quantitative metrics directly related to the child's performance.
As an efficient approach to understand, generate, and process natural language texts, research in natural language processing (NLP) has exhibited a rapid spread and wide adoption in recent years. Given the increasing research work in this area, several NLP-related approaches have been surveyed in the research community. However, a comprehensive study that categorizes established topics, identifies trends, and outlines areas for future research remains absent. Contributing to closing this gap, we have systematically classified and analyzed research papers in the ACL Anthology. As a result, we present a structured overview of the research landscape, provide a taxonomy of fields of study in NLP, analyze recent developments in NLP, summarize our findings, and highlight directions for future work.
Bernadette Witecy, Eva Wimmer, Isabel Neitzel
et al.
IntroductionThe present study provides longitudinal data on the development of receptive and expressive grammar in children and adolescents with Down syndrome and addresses the role of nonverbal cognitive abilities and verbal short-term memory for morphosyntactic development.MethodSeventeen German-speaking individuals with Down syndrome (aged 4;6–17;1 years at first testing (T1)) were assessed twice, 4;4–6;6 years apart. For a subset of five participants, there was also a third assessment 2 years after the second. Receptive grammar, nonverbal cognition, and verbal short-term memory were tested using standardized measures. For expressive grammar, elicitation tasks were used to assess the production of subject-verb agreement and of wh-questions.ResultsAt group level, the participants showed a significant increase in grammar comprehension from T1 to T2. However, progress diminished with increasing chronological age. Notable growth could not be observed beyond the age of 10 years.With respect to expressive grammatical abilities, progress was limited to those participants who had mastered verbal agreement inflection around age 10 years. Individuals who did not master verbal agreement by late childhood achieved no progress in producing wh-questions, either.There was an increase in nonverbal cognitive abilities in the majority of participants. Results for verbal short-term memory followed a similar pattern as those for grammar comprehension. Finally, neither nonverbal cognition nor verbal short-term memory were related to changes in receptive or expressive grammar.DiscussionThe results point to a slowdown in the acquisition of receptive grammar which starts before the teenage years. For expressive grammar, improvement in wh-question production only occurred in individuals with good performance in subject-verb agreement marking, which suggests that the latter might have a trigger function for further grammatical development in German-speaking individuals with Down syndrome. The study provides no indication that nonverbal cognitive abilities or verbal short-term memory performance determined the receptive or expressive development. The results lead to clinical implications for language therapy.
Since English is a subject that all students in Indonesian schools are required to study, schools are under pressure to make studying the language enjoyable for the pupils. The use of the Duolingo app to increase the English vocabulary of Grade 3 A MI Tarbiyatul Islam pupils is discussed in this study. memorizing a language involves memorizing a lot of vocabulary. This study employs a qualitative case study methodology. Four people the head of the madrasa, an English teacher, and two Class 3 A students from MI Tarbiyatul Islam served as the study's informants. Data gathering, data reduction, data display, and conclusions are all used in data analysis. According to the research findings, the Duolingo application can increase students' vocabulary acquisition because of its supporting elements, which include an initial ability analysis, engaging the students' attention, repetition, and practice questions.
Marcel Schlechtweg, Jörg Peters, Marina Frank
et al.
A person’s first language (L1) affects the way they acquire speech in a second language (L2). However, we know relatively little about the role different varieties of the L1 play in the acquisition of L2 speech. This study focuses on German (L1) learners of English (L2) and asks whether the degree to which German speakers distinguish between the two vowels /eː/ and /ɛː/ in their L1 has an impact on how well these individuals identify /æ/ and discriminate between the two English vowels /ɛ/ and /æ/. These two English vowels differ in both vowel quality and duration (/æ/ is longer than /ɛ/). We report on an identification and a discrimination experiment. In the first study, participants heard a sound file and were asked to indicate whether they heard “pen” or “pan” (or “pedal” or “paddle”). The stimuli differed from each other in terms of both vowel quality (11 steps on a spectral continuum from an extreme /æ/ to an extreme /ɛ/) and duration (short, middle, long). In the second study, participants had to signal whether two sound files they were exposed to differed from each other. We modeled the percentage of /æ/ (“pan,” “paddle”) selection (identification task only, binomial logistic regression), accuracy (discrimination task only, binomial logistic regression), and reaction time (identification and discrimination tasks, linear mixed effects models) by implementing the German Pillai score as a measure of vowel overlap in our analysis. Each participant has an individual Pillai score, which ranges from 0 (= merger of L1 German /eː/ and /ɛː/) to 1 (=maintenance of the contrast between L1 German /eː/ and /ɛː/) and had been established, prior to the perception experiments reported here, in a production study. Although the findings from the discrimination study remain inconclusive, the results from the identification test support the hypothesis that maintaining the vowel contrast in the L1 German leads to a more native-like identification of /æ/ in L2 English. We conclude that sociolinguistic variation in someone’s L1 can affect L2 acquisition.
AbstractSLA research is characterised by a striking homogeneity in the linguistic, social and geographical data we as a field draw on. Such empirical homogeneity is a potential threat to the validity and scope of our models and theories. This paper focuses on a particular gap in our knowledge, namely the SLA of sign languages. It outlines an argument as to why the SLA of sign matters to general SLA research in terms of the empirical representativity, generalisability, and validity of the conclusions in the field. It exemplifies three domains where the study of language acquisition across modalities could shed important light on theoretical issues in mainstream SLA/bilingualism research (e.g. learner varieties, explicit-implicit learning, and crosslinguistic influence), and highlight some of the methodological challenges involved in such work.
We introduce GPT-NeoX-20B, a 20 billion parameter autoregressive language model trained on the Pile, whose weights will be made freely and openly available to the public through a permissive license. It is, to the best of our knowledge, the largest dense autoregressive model that has publicly available weights at the time of submission. In this work, we describe \model{}'s architecture and training and evaluate its performance on a range of language-understanding, mathematics, and knowledge-based tasks. We find that GPT-NeoX-20B is a particularly powerful few-shot reasoner and gains far more in performance when evaluated five-shot than similarly sized GPT-3 and FairSeq models. We open-source the training and evaluation code, as well as the model weights, at https://github.com/EleutherAI/gpt-neox.
Scandinavian countries are perceived as role-models when it comes to gender equality. With the advent of pre-trained language models and their widespread usage, we investigate to what extent gender-based harmful and toxic content exist in selected Scandinavian language models. We examine nine models, covering Danish, Swedish, and Norwegian, by manually creating template-based sentences and probing the models for completion. We evaluate the completions using two methods for measuring harmful and toxic completions and provide a thorough analysis of the results. We show that Scandinavian pre-trained language models contain harmful and gender-based stereotypes with similar values across all languages. This finding goes against the general expectations related to gender equality in Scandinavian countries and shows the possible problematic outcomes of using such models in real-world settings.
Due to the considerable increase in freely available data (especially on the Web), extracting relevant information from textual content is a critical challenge. Most of the available data is embedded in unstructured texts and is not linked to formalized knowledge structures such as ontologies or rules. A potential solution to this problem is to acquire such knowledge through natural language processing (NLP) tools and text mining techniques. Prior work has focused on the automatic extraction of ontologies from texts, but the acquired knowledge is generally limited to simple hierarchies of terms. This paper presents a polyvalent framework for acquiring complex relationships from texts and coding these in the form of rules. Our approach begins with existing domain knowledge represented as an OWL ontology, and applies NLP tools and text matching techniques to deduce different atoms, such as classes, properties and literals, to capture deductive knowledge in the form of new rules. For the reason, to enrich the existing domain ontology by these rules, in order to obtain higher relational expressiveness, make reasoning and produce new facts. The approach was tested using medical reports, specifically, in the specialty of gynecology. It reports an F-measure of 95.83% on test our corpus.
Adult language learners show distinct abilities in acquiring a new language, yet the underlying neural mechanisms remain elusive. Previous studies suggested that resting-state brain connectome may contribute to individual differences in learning ability. Here, we recorded electroencephalography (EEG) in a large cohort of 106 healthy young adults (50 males) and examined the associations between resting-state alpha band (8–12 Hz) connectome and individual learning ability during novel word learning, a key component of new language acquisition. Behavioral data revealed robust individual differences in the performance of the novel word learning task, which correlated with their performance in the language aptitude test. EEG data showed that individual resting-state alpha band coherence between occipital and frontal regions positively correlated with differential word learning performance (p = 0.001). The significant positive correlations between resting-state occipito-frontal alpha connectome and differential world learning ability were replicated in an independent cohort of 35 healthy adults. These findings support the key role of occipito-frontal network in novel word learning and suggest that resting-state EEG connectome may be a reliable marker for individual ability during new language learning.
The present study is a longitudinal study for approximately 26 months to the Indonesian child and has been through her second language acquisition in Japan. A Longitudinal study is a research design that involved repeated observation of the same variables over long periods. The acquisition process took place for about four years. After returning to Indonesia, the family wants to keep her second language and do some second language maintenance. While in her process to be bilingual, she experienced a process of code-switching and code-mixing in her daily life using their mother tongue, Indonesian, and her second language, Japanese. This research focuses on how the child maintains her second language and how the bilingual process's phenomena occur through interactions in the family environment. Several language transfers from the second language to the first language occur in their daily life using Indonesian. This study uses an ethnographic research approach. Conducting ethnographic research requires a long-term process by making detailed notes about the group's behavior and beliefs from time to time. Observation and interviews are the procedures used in data collection in the field. The transfer language process is used through the code-mixing, code-switching, and preservation process of the second language after returning home. The results saw that the child both uses language systems in each language and sometimes mixed in between languages, as she has her languages.
Keywords: code-switching; language mixing; Japanese as a second language; bilingual process
English language, Language. Linguistic theory. Comparative grammar
The paper explores the origin of a foreign accent in non-native speech. Deviations from the pronun-ciation norm in the articulation of English vowels and consonants by Ukrainian learners of English form a dynamic system of specific features correlating with the degree of foreign language competence. Phonetic interference of native language production and perception habits into the foreign language performance has its psychological and linguistic reasons. Divergence of phonological and phonetic features of native and foreign languages, automated articulations transferred into the foreign speech shape the specific character of the foreign accent. The contrastive analysis of the articulatory bases of English and Ukrainian as well as the analysis of actual phonetic deviations enabled to single out salient features of Ukrainian English accent.
Im Beitrag wird die Herkunft eines fremden nicht-muttersprachlichen Akzents untersucht. Abweichungen von der Aussprachenorm bei der Artikulation von englischen Vokalen und Konsonanten durch ukrainische Englischlerner bilden ein dynamisches System von spezifischen Merkmalen, die mit dem Grad der Fremd-sprachenkompetenz zusammenhängen. Die phonetische Interferenz der muttersprachlichen Produktions – und Wahrnehmungsgewohnheiten auf die fremdsprachlichen Leistungen hat psychologische und sprach-liche Gründe. Die Divergenz der phonetischen und phonologischen Merkmale der Muttersprache und der Fremdsprache sowie die automatisierten Artikulationen, die in die Fremdsprache übertragen werden, prägen den spezifischen Charakter des fremden Akzents. Die kontrastive Analyse der Artikulationsgrundlagen des Englischen und Ukrainischen sowie die Analyse der tatsächlichen phonetischen Abweichungen ermöglichten es, herausragende Merkmale des ukrainischen Englischakzents herauszustellen.
Irina A. Novikova, Nadezhda S. Berisha, Alexey L. Novikov
et al.
Foreign (second) language (FL/SL) proficiency is one of the most important competencies for a modern person, and is necessary for both professional and personal fulfillment. The purpose of this study is to consider and compare personality traits and creativity as predictors of success in foreign language acquisition (FLA). The sample includes 128 (105 female and 23 male) first- and second-year university linguistics students. Creativity is measured by the Abbreviated Torrance Test for Adults (ATTA). The FFM personality traits are measured by the Russian NEO Five-Factor Inventory adaptation by S. Biryukov and M. Bodunov. To assess the level of FL proficiency, we used a traditional academic achievement indicator (the semester’s final grades in English), as well as the English teachers’ assessment of the level of language proficiency of their students according to the “Foreign Language Proficiency Scale” (10 indicators and total score). Descriptive statistics methods and a multiple regression analysis were used to process the data in the R software environment, version 3.5.2. The findings of our research showed that creativity indicators have a stronger but contradictory impact on the level of FL proficiency compared to personality traits. We suggest that teachers, most likely, lack knowledge on the manifestations of student creativity in the process of FL learning.
In this work, we focus on the problem of grounding language by training an agent to follow a set of natural language instructions and navigate to a target object in an environment. The agent receives visual information through raw pixels and a natural language instruction telling what task needs to be achieved and is trained in an end-to-end way. We develop an attention mechanism for multi-modal fusion of visual and textual modalities that allows the agent to learn to complete the task and achieve language grounding. Our experimental results show that our attention mechanism outperforms the existing multi-modal fusion mechanisms proposed for both 2D and 3D environments in order to solve the above-mentioned task in terms of both speed and success rate. We show that the learnt textual representations are semantically meaningful as they follow vector arithmetic in the embedding space. The effectiveness of our attention approach over the contemporary fusion mechanisms is also highlighted from the textual embeddings learnt by the different approaches. We also show that our model generalizes effectively to unseen scenarios and exhibit zero-shot generalization capabilities both in 2D and 3D environments. The code for our 2D environment as well as the models that we developed for both 2D and 3D are available at https://github.com/rl-lang-grounding/rl-lang-ground.