Papers reviewed and determined not to be word norm studies. Use the flag icon to report errors or suggest re-inclusion.
18625 papers
The outstanding performance recently reached by neural language models (NLMs) across many natural language processing (NLP) tasks has steered the debate towards understanding whether NLMs implicitly learn linguistic competence. Probes, i.e., supervised models trained using NLM representations to predict linguistic properties, are frequently adopted to investigate this issue. However, it is still questioned if probing classification tasks really enable such investigation or if they simply hint at surface patterns in the data. This work contributes to this debate by presenting an approach to assessing the effectiveness of a suite of probing tasks aimed at testing the linguistic knowledge implicitly encoded by one of the most prominent NLMs, BERT. To this aim, we compared the performance of probes when predicting gold and automatically altered values of a set of linguistic features. Our experiments were performed on Italian and were evaluated across BERT’s layers and for sentences with different lengths. As a general result, we observed higher performance in the prediction of gold values, thus suggesting that the probing model is sensitive to the distortion of feature values. However, our experiments also showed that the length of a sentence is a highly influential factor that is able to confound the probing model’s predictions.
Social distance, or perception of the other, is recognized as a dynamic dimension of an interaction, but yet to be widely explored or understood. Through CORAE, a novel web-based open-source tool for COntinuous Retrospective Affect Evaluation, we collected retrospective ratings of interpersonal perceptions between 12 participant dyads. In this work, we explore how different aspects of these interactions reflect on the ratings collected, through a discourse analysis of individual and social behavior of the interactants. We found that different events observed in the ratings can be mapped to complex interaction phenomena, shedding light on relevant interaction features that may play a role in interpersonal understanding and grounding. This paves the way for better, more seamless human-robot interactions, where affect is interpreted as highly dynamic and contingent on interaction history.
L1-L2 parallel dependency treebanks are UD-annotated corpora of learner sentences paired with correction hypotheses. Automatic morphosyntactical annotation has the potential to remove the need for explicit manual error tagging and improve interoperability, but makes it more challenging to locate grammatical errors in the resulting datasets. We therefore propose a novel method for automatically extracting morphosyntactical error patterns and perform a preliminary bilingual evaluation of its first implementation through a similar example retrieval task. The resulting pipeline is also available as a prototype CALL application.
Recent efforts to consolidate guidelines and treebanks in the Universal Dependencies project raise the expectation that joint training and dataset comparison is increasingly possible for high-resource languages such as English, which have multiple corpora. Focusing on the two largest UD English treebanks, we examine progress in data consolidation and answer several questions: Are UD English treebanks becoming more internally consistent? Are they becoming more like each other and to what extent? Is joint training a good idea, and if so, since which UD version? Our results indicate that while consolidation has made progress, joint models may still suffer from inconsistencies, which hamper their ability to leverage a larger pool of training data.
BACKGROUND: Psychiatric comorbidities are common in Parkinson's disease (PD) and may change with high-frequency stimulation targeting the subthalamic nucleus. Numerous accounts indicate subthalamic alpha-frequency oscillation is implicated in emotional processing. While intermittent alpha-frequency (10Hz) stimulation induces positive emotional effects, with more ventromedial contacts inducing larger effects, little is known about the subacute effect of ventral 10Hz subthalamic stimulation on emotional processing. OBJECTIVE/HYPOTHESIS: To evaluate the subacute effect of 10Hz stimulation at bilateral ventral subthalamic nucleus on emotional processing in PD patients using an affective task, compared to that of clinical-frequency stimulation and off-stimulation. METHODS: Twenty PD patients with bilateral subthalamic deep brain stimulation for more than six months were tested with the affective task under three stimulation conditions (10Hz, 130Hz, and off-stimulation) in a double-blinded randomized design. RESULTS: While 130Hz stimulation reduced arousal ratings in all patients, 10Hz stimulation increased arousal selectively in patients with higher depression scores. Furthermore, 10Hz stimulation induced a positive shift in valence rating to negative emotional stimuli in patients with lower apathy scores, and 130Hz stimulation led to more positive valence to emotional stimuli in the patients with higher apathy scores. Notably, we found correlational relationships between stimulation site and affective rating: arousal ratings increase with stimulation from anterior to posterior site, and positive valence ratings increase with stimulation from dorsal to ventral site of the ventral subthalamic nucleus. CONCLUSIONS: Our findings highlight the distinctive role of 10Hz stimulation on subjective emotional experience and unveil the spatial organization of the stimulation effect.
In contrast with findings of reduced facial and vocal emotional recognition (ER) accuracy, children on the autism spectrum (AS) demonstrate comparable ER skills to those of typically-developing (TD) children using music. To understand the specificity of purported ER differences, the goal of this study was to examine ER from music compared with faces and voices among children on the AS and TD children. Twenty-five children on the AS and 23 TD children (6-13 years) completed an ER task, using categorical (happy, sad, fear) and dimensional (valence, arousal) ratings, of emotions presented via music, faces, or voices. Compared to the TD group, the AS group showed a relative ER strength from music, and comparable performance from faces and voices. Although both groups demonstrated greater vocal ER accuracy, the children on the AS performed equally well with music and faces, whereas the TD children performed better with faces than with music. Both groups performed comparably with dimensional ratings, except for greater variability by the children on the AS in valence ratings for happy emotions. These findings highlight a need to re-examine ER of children on the AS, and to consider how facilitating strengths-based approaches can re-shape our thinking about and support for persons on the AS.
Abstract This paper presents the introduction of frame semantics in SciE-Lex, a lexical database of biomedical English, in order to establish frame-based semantic networks among the lexical units contained in the database and draw attention to the domain-specific meanings and syntactic patterns that they exhibit. Taking the general English FrameNet database as a reference and the verbs diminish, compete, and perturb as an illustration, this paper shows how modelling the syntacticosemantic properties of biomedical English results in three scenarios: (a) the general English FrameNet analysis is fully adequate for biomedical usage (b) the frame elements in the general English FrameNet do not fully characterize the semantic and syntactic properties of the lexical items in biomedical English, so an existing frame needs to be customized (c) the frame(s) evoked in general English are different, so either the customization of an existing frame or the addition of a new frame is necessary.
This paper aims to provide clear guidelines for researchers studying conspiracy theory belief. It examines the meta-linguistic question about how we should conceptaulize 'conspiracy theory' and its relationship to the evaluative question of how we should evaluate beliefs in conspiracy theories, addressing normative issues surrounding the meaning, use, and conceptualization of ‘conspiracy theory’, as well as how these issues might impact how researchers study conspiracy theories or beliefs in them It argues that four norms, the Empirical Accuracy Norm, the Linguistic Norm, the Social Norm, and the Academic Fecundity Norm, underlie debates about how we should conceptualize or define ‘conspiracy theory’. We zoom in on the linguistic norm, as it has been treated as more fundamental than the other norms. We then scrutinize the argument that normative conceptualizations prematurely settle the question of how conspiracy theories and belief in them should be evaluated, and argue that it fails. Subsequently, we turn to the risks normative conceptualizations pose when it comes to certain assumptions and biases in the study of conspiracy theory belief. Finally, we explore where this leaves us regarding the meta-linguistic and evaluative questions, and formulate seven guidelines for studying conspiracy theory belief, whether it be theoretical, historical, or empirical.
Purpose: This study aims to examine the impact of sleep deprivation on individual cognitive reappraisal ability using a standardized behavioral paradigm. Methods: A randomized pretest-posttest control group design was conducted. Thirty-nine participants were eventually enrolled and randomly assigned to receive either the sleep control (SC: n = 17) or the sleep deprivation (SD: n = 22). Both of them were required to perform a standardized behavioral paradigm of measuring cognitive reappraisal ability one time under sleep-rested condition and another time under the condition of different sleep manipulation a week later. Results: Mean valence ratings of SD group were more negative than SC group's (p < 0.05) and mean arousal ratings of SD group were higher than SC group's (p < 0.01). Conclusion: Sleep deprivation may impair individual cognitive reappraisal ability and could potentially undermine the efficacy of cognitive therapy in terms of emotion regulation.
The early posterior negativity (EPN) is a mid-latency ERP component that is reliably enhanced by emotional cues, with a deflection beginning between 150 and 200 msec after stimulus onset. The brief, bilateral occipital EPN is followed by the centroparietal late positive potential (LPP), a long duration slow-wave that is strongly associated with emotional arousal ratings of scenes. A recent study suggests that the EPN is particularly sensitive to human bodies in scenes, independent of emotional intensity. Here, we directly investigate the influence of human body features on EPN modulation, using emotional and neutral scenes depicting people across a range of body exposures and orientations, in addition to scenes of pleasant, neutral, and unpleasant animals. The results demonstrate that the EPN is quite sensitive to human body features and weakly related to arousal ratings, whereas the LPP is strongly modulated by scenes that receive high arousal ratings. Based on these results and relevant work on body-specific visual perception, we speculate that modulation of the EPN may strongly reflect the early detection of human bodies, which serves as a predictor of emotional significance, whereas LPP modulation is more closely associated with the extended elaborative processing of scenes that are explicitly judged to be emotionally arousing.
Abstract In this study, we compared affective ratings of emotional valence and arousal for 882 Serbian words at three points in time: before the onset of the COVID-19 pandemic (2018), during the COVID-19 lockdown (2020), and after the government measures were abandoned (2022). We did not observe a significant change in average valence or arousal ratings across time points. A more detailed look into the data revealed the change in arousal that was different across the valence values. An increase in their linear correlations and a decrease in the nonlinearity of the GAMM smooth demonstrated that, upon the onset of the COVID-19 pandemic, emotionally negative words elicited higher arousal ratings, whereas emotionally positive words elicited lower arousal ratings. It revealed that our participants became more sensitive to the negative content and less sensitive to the positive content. Our results add to the findings, which showed that the relationship between emotional valence and arousal is a function of contextual factors, which primarily influence the arousal of words.
With the advent of pre-trained language models, many natural language processing tasks in various languages have achieved great success. Although some research has been conducted on fine-tuning BERT-based models for syntactic parsing, and several Arabic pre-trained models have been developed, no attention has been paid to Arabic dependency parsing. In this study, we attempt to fill this gap and compare nine Arabic models, fine-tuning strategies, and encoding methods for dependency parsing. We evaluated three treebanks to highlight the best options and methods for fine-tuning Arabic BERT-based models to capture syntactic dependencies in the data. Our exploratory results show that the AraBERTv2 model provides the best scores for all treebanks and confirm that fine-tuning to the higher layers of pre-trained models is required. However, adding additional neural network layers to those models drops the accuracy. Additionally, we found that the treebanks have differences in the encoding techniques that give the highest scores. The analysis of the errors obtained by the test examples highlights four issues that have an important effect on the results: parse tree post-processing, contextualized embeddings, erroneous tokenization, and erroneous annotation. This study reveals a direction for future research to achieve enhanced Arabic BERT-based syntactic parsing.
Aesthetic chills are an embodied peak emotional experience induced by stimuli such as music, films, and speeches and characterized by dopaminergic release. The emotional consequences of chills in terms of valence and arousal are still debated and the existing empirical data is conflicting. In this study, we tested the effects of ChillsDB, an open-source repository of chills-inducing stimuli, on the emotional ratings of 600+ participants. We found that participants experiencing chills reported significantly more positive valence and greater arousal during the experience, compared to participants who did not experience chills. This suggests that the embodied experience of chills may influence one's perception and affective evaluation of the context, in favor of theoretical models emphasizing the role of interoceptive signals such as chills in the process of perception and decision-making. We also found an interesting pattern in the valence ratings of participants, which tended to harmonize toward a similar mean after the experiment, though initially disparately distributed. We discuss the significance of these results for the diagnosis and treatment of dopaminergic disorders such as Parkinson's, schizophrenia, and depression.
Recursive Neural Networks (RNNs) are a powerful tool for Natural Language Processing (NLP) that can easily learn and represent recursive structures in sentences. Artificial Neural Networks (ANNs) have been widely used in various NLP tasks, such as language modeling, text classification, and sentiment analysis. With the development of deep learning, RNNs have emerged as a powerful tool to represent structural and syntactic information of sentences. RNNs have attracted a lot of attention due to their ability to learn complex language structures and representation. Consequently, RNNs have achieved great performance in a variety of NLP applications, such as text summarization, machine translation, and question answering. In this paper, we discuss the importance of RNNs in NLP and present different smart performance optimization strategies for RNNs. Firstly, we consider the use of pre-trained language models such as GloVe and word2vec to enhance the representation of words and improve the overall accuracy. Secondly, we implement a variety of regularization techniques, such as dropout and batch normalization, to regularize the learning process of RNNs. Finally, we discuss the approaches used to reduce the complexity and overfitting of RNNs. We compare the performance of different RNNs on the Penn Treebank dataset with different optimization strategies. Experimental results show that our proposed approaches can significantly improve the performance of RNNs and provide a more effective solution for NLP tasks.
Useful information from natural language can be extracted by the process of natural language processing (NLP). NLP contains various tools where parts-of-speech (PoS) tagging plays a significant part. Every word in a sentence can be tagged by the process of PoS which includes noun, adjective, preposition, pronoun, article, verb, adverb and many more. In Penn Treebank have a 48 tagset, so we need to add many more tag sets. Penn Treebank in its eight years of procedures composed of generally seven trillion part-of-speech words tagging markers, three trillion words of underfed translate contents, over two trillion words of content translate for state words format, and 1.6 trillion sentences reproduced verbal content observation for oral communication speech. All these materials consist of such huge genres, for example, IBM PC manuals and WSJ (Wall Street Journal) and so on. This paper defines a review of various techniques, for PoS tagging and the PoS tagging process.
INTRODUCTION: Craving, involving intense and urgent desires to engage in specific behaviours, is a feature of addictions. Multiple studies implicate regions of salience/limbic networks and basal ganglia, fronto-parietal, medial frontal regions in craving in addictions. However, prior studies have not identified common neural networks that reliably predict craving across substance and behavioural addictions. METHODS: Functional magnetic resonance imaging during an audiovisual cue-reactivity task and connectome-based predictive modelling (CPM), a data-driven method for generating brain-behavioural models, were used to study individuals with cocaine-use disorder and gambling disorder. Functions of nodes and networks relevant to craving were identified and interpreted based on meta-analytic data. RESULTS: Craving was predicted by neural connectivity across disorders. The highest degree nodes were mostly located in the prefrontal cortex. Overall, the prediction model included complex networks including motor/sensory, fronto-parietal, and default-mode networks. The decoding revealed high functional associations with components of memory, valence ratings, physiological responses, and finger movement/motor imagery. CONCLUSIONS: Craving could be predicted across substance and behavioural addictions. The model may reflect general neural mechanisms of craving despite specificities of individual disorders. Prefrontal regions associated with working memory and autobiographical memory seem important in predicting craving. For further validation, the model should be tested in diverse samples and contexts.
Constituency parsing is an important task of informing how words are combined to form sentences. While constituency parsing in English has seen significant progress in the last few years, tools for constituency parsing in Indonesian remain few and far between. In this work, we publish ICON (Indonesian CONstituency treebank), the hitherto largest publicly available manually-annotated benchmark Indonesian constituency treebank with a size of 10,000 sentences and approximately 124,000 constituents and 182,000 tokens, which can support the training of state-of-the-art transformer-based models. As part of the process of building the treebank, we review and revamp the constituent and POS tagsets in use in existing treebanks to ensure that the labels are relevant and suitable for the grammatical features of Indonesian. We establish strong baselines on the ICON dataset using the Berkeley Neural Parser with transformer-based pre-trained embeddings, with the best performance of 88.85% F1 score coming from our own version of SpanBERT (IndoSpanBERT). We further analyze the predictions made by our best-performing model to reveal certain idiosyncrasies in Indonesian that pose challenges for constituency parsing.
From the corpus data, we observe that in the real language usage, the particular verb does not appear in all theoretically possible finite and infinite verb forms in the morphologically rich Lithuanian but is used in those forms which are relevant for the verb patterning. On the one hand, by teaching vocabulary, is it important to represent lexis in these relevant forms – frequently used forms, and, on the other hand, in grammar teaching, there is a need to provide learners with appropriate vocabulary, e.g., by teaching infinite forms, to use verbs, in the usage of which, these forms are relevant and frequent.In this paper, we provide language teaching practitioners with the data about the frequently used Lithuanian verbs and show which of them and how often appear in infinite forms (participles in passive and active voice, adverbial participles, half participles). As a research data we use 200 verbs from the Lexical Database of Lithuanian Language Usage which was developed on the basis of the written subcorpus of the Pedagogic corpus of Lithuanian. The investigated verbs belong to the frequent vocabulary: in the corpus of approx. 700,000 tokens, these verbs are used 100 times (and above). First, we analysed, which verbs appear in infinite forms, second, we checked whether frequent and typical infinite forms are included into corpus pattern(s) of these particular verbs, and if there is a link between the infinite form and a particular meaning of the verb.All verbs (except of three verbs with no infinite forms) were included into one of three groups: 1) 11 verbs which occur in the infinite forms frequently (more than 50% of all forms – finite and infinite) and, accordingly, typical; 2) 117 verbs with the infinite forms making up from 10 to 50%; 3) 69 verbs, with the infinite forms making up less than 10% of all verb forms. Interestingly, the verbs of the first group, usually have only one infinite form, e.g., participle in passive voice which makes up more than 50% of all forms of verb. These cases are also frequently observed in the second verb group. Thus, if the verb tends to be used in infinite forms, it is important to know which infinite form is relevant to that particular verb.In the Lexical Database of Lithuanian Language Usage, lexical and grammatical patterning of the word is represented in the form of corpus patterns. In this study, we showed the interrelation between the frequently used infinite forms of the verb and its corpus patterns (also, corpus patterns related to particular meaning of the polysemous verb). We can expect various applications of the provided data in the Lithuanian as a foreign language teaching: the provided data about the verbs typical and frequent in infinite forms and the corpus patterns including these infinite forms can be used for building vocabulary training as well as for developing grammar exercises.
Abstract Viewing a live facial expression typically elicits a similar expression by the observer (facial mimicry) that is associated with a concordant emotional experience (emotional contagion). The model of embodied emotion proposes that emotional contagion and facial mimicry are functionally linked although the neural underpinnings are not known. To address this knowledge gap, we employed a live two-person paradigm (n = 20 dyads) using functional near-infrared spectroscopy during live emotive face-processing while also measuring eye-tracking, facial classifications and ratings of emotion. One dyadic partner, ‘Movie Watcher’, was instructed to emote natural facial expressions while viewing evocative short movie clips. The other dyadic partner, ‘Face Watcher’, viewed the Movie Watcher's face. Task and rest blocks were implemented by timed epochs of clear and opaque glass that separated partners. Dyadic roles were alternated during the experiment. Mean cross-partner correlations of facial expressions (r = 0.36 ± 0.11 s.e.m.) and mean cross-partner affect ratings (r = 0.67 ± 0.04) were consistent with facial mimicry and emotional contagion, respectively. Neural correlates of emotional contagion based on covariates of partner affect ratings included angular and supramarginal gyri, whereas neural correlates of the live facial action units included motor cortex and ventral face-processing areas. Findings suggest distinct neural components for facial mimicry and emotional contagion. This article is part of a discussion meeting issue ‘Face2face: advancing the science of social interaction’.
Anthropomorphic appearance is a key factor to affect users’ attitudes and emotions. This research aimed to measure emotional experience caused by robots’ anthropomorphic appearance with three levels – high, moderate, and low – using multimodal measurement. Fifty participants’ physiological and eye-tracker data were recorded synchronously while they observed robot images that were displayed in random order. Afterward, the participants reported subjective emotional experiences and attitudes towards those robots. The results showed that the images of the moderately anthropomorphic service robots induced higher pleasure and arousal ratings, and yielded significantly larger pupil diameter and faster saccade velocity, than did the low or high robots. Moreover, participants’ facial electromyography, skin conductance, and heart-rate responses were higher when observing moderately anthropomorphic service robots. An implication of the research is that service robots’ appearance should be designed to be moderately anthropomorphic; too many human-like features or machine-like features may disturb users’ positive emotions and attitudes.Practitioner Summary: This research aimed to measure emotional experience caused by three types of anthropomorphic service robots using a multimodal measurement experiment. The results showed that moderately anthropomorphic service robots evoked more positive emotion than high and low anthropomorphic robots. Too many human-like features or machine-like features may disturb users’ positive emotions.
This paper demonstrates the development and evaluation of a multilingual neural machine translation system for Indian languages based on the mT5 transformer, successfully utilized to develop multiple state-of-the-art NLP models. We used the modified Asian Language Treebank multilingual dataset to train the system for developing a Machine Translation model capable of translating text in English, Hindi and Bengali amongst each other. Our system was able to achieve acceptable BLEU scores of over 20 in five of the six language pairs, with the English to Bengali system achieving a maximum BLEU score of 49.87 and the Bengali to English system achieving an average BLEU score of 42.43.
Abstract Gender can be considered an embodied social concept encompassing biological and cultural components. In this study, we explored whether the concept of gender varies as a function of different cultural and linguistic norms by comparing communities that vary in their social treatment of gender-related issues and linguistic encoding of gender. In Study 1, Italian, Dutch, and English-speaking participants completed a free-listing task, which showed Italians and Dutch were the most distinct in their conceptualization of gender: Italian participants focused more on socio-cultural features (e.g., discrimination, politics, and power ), whereas Dutch participants focused more on the corporeal sphere (e.g., hormones, breasts, and genitals ). Study 2 replicated this finding focusing on Italian and Dutch and using a typicality rating task: socio-cultural and abstract features were considered as more typical of “gender” by Italian than Dutch participants. Study 3 addressed Italian and Dutch participants’ explicit beliefs about gender with a questionnaire measuring essentialism and constructivism, and consolidated results from Studies 1 and 2 showing that Dutch participants endorsed more essentialist beliefs about gender than Italian participants. Consistent with socio-cultural constructivist accounts, our results provide evidence that gender is conceptualized differently by diverse groups and is adapted to specific cultural and linguistic environments.
Feeling happy, or judging whether someone else is feeling happy are two distinct facets of emotions that nevertheless rely on similar physiological and neural activity. Differentiating between these two states, also called Self/Other distinction, is an essential aspect of empathy, but how exactly is it implemented? In non-emotional cognition, the transient neural response evoked at each heartbeat, or heartbeat evoked response (HER), indexes the self and signals Self/Other distinction. Here, using electroencephalography (n = 32), we probe whether HERs' role in Self/Other distinction extends also to emotion - a domain where brain-body interactions are particularly relevant. We asked participants to rate independently validated affective scenes, reporting either their own emotion (Self) or the emotion expressed by people in the scene (Other). During the visual cue indicating to adopt the Self or Other perspective, before the affective scene, HERs distinguished between the two conditions, in visual cortices as well as in the right frontal operculum. Physiological reactivity (facial electromyogram, skin conductance, heart rate) during affective scene co-varied as expected with valence and arousal ratings, but also with the Self- or Other- perspective adopted. Finally, HERs contributed to the subjective experience of valence in the Self condition, in addition to and independently from physiological reactivity. We thus show that HERs represent a trans-domain marker of Self/Other distinction, here specifically contributing to experienced valence. We propose that HERs represent a form of evidence related to the 'I' part of the judgement 'To which extent do I feel happy'. The 'I' related evidence would be combined with the affective evidence collected during affective scene presentation, accounting at least partly for the difference between feeling an emotion and identifying it in someone else.
Listening to music is an enjoyable behaviour that engages multiple networks of brain regions. As such, the act of music listening may offer a way to interrogate network activity, and to examine the reconfigurations of brain networks that have been observed in healthy aging. The present study is an exploratory examination of brain network dynamics during music listening in healthy older and younger adults. Network measures were extracted and analyzed together with behavioural data using a combination of hidden Markov modelling and partial least squares. We found age- and preference-related differences in fMRI data collected during music listening in healthy younger and older adults. Both age groups showed higher occupancy (the proportion of time a network was active) in a temporal-mesolimbic network while listening to self-selected music. Activity in this network was strongly positively correlated with liking and familiarity ratings in younger adults, but less so in older adults. Additionally, older adults showed a higher degree of correlation between liking and familiarity ratings consistent with past behavioural work on age-related dedifferentiation. We conclude that, while older adults do show network and behaviour patterns consistent with dedifferentiation, activity in the temporal-mesolimbic network is relatively robust to dedifferentiation. These findings may help explain how music listening remains meaningful and rewarding in old age.
In this paper we present a novel treebank developed to analyse marked constructions in Italian called MarkIT.The resource contains almost 1,300 sentences manually annotated with dependency relations following the Universal Dependencies paradigm.The sentences have been extracted from essays written by high-school students along several years, which accounts for the structure and the topic variability of the sentences.In this work, we detail the process to select the sentences, parse them automatically and then manually correct them.The resource covers seven types of marked constructions (839 sentences overall) plus some sentences, whose syntax can be wrongly classified as marked and which can serve as negative examples of markedness (453 sentences).We also present an evaluation of parsing performance, comparing a model trained on existing Italian treebanks with the model obtained by adding MarkIT to the training set.
Perception of passing time can be distorted. 1 Emotional experiences, particularly arousal, can contract or expand experienced duration via their interactions with attentional and sensory processing mechanisms. 2, 3 Current models suggest that perceived duration can be encoded from accumulation processes 4, 5 and from temporally evolving neural dynamics. 6, 7 Yet all neural dynamics and information processing ensue at the backdrop of continuous interoceptive signals originating from within the body. Indeed, phasic fluctuations within the cardiac cycle impact neural and information processing. 8, 9, 10, 11, 12, 13, 14, 15 Here, we show that these momentary cardiac fluctuations distort experienced time and that their effect interacts with subjectively experienced arousal. In a temporal bisection task, durations (200–400 ms) of an emotionally neutral visual shape or auditory tone (experiment 1) or of an image displaying happy or fearful facial expressions (experiment 2) were categorized as short or long. 16 Across both experiments, stimulus presentation was time-locked to systole, when the heart contracts and baroreceptors fire signals to the brain, and to diastole, when the heart relaxes, and baroreceptors are quiescent. When participants judged the duration of emotionally neural stimuli (experiment 1), systole led to temporal contraction, whereas diastole led to temporal expansion. Such cardiac-led distortions were further modulated by the arousal ratings of the perceived facial expressions (experiment 2). At low arousal, systole contracted while diastole expanded time, but as arousal increased, this cardiac-led time distortion disappeared, shifting duration perception toward contraction. Thus, experienced time contracts and expands within each heartbeat—a balance that is disrupted under heightened arousal.
Analiza glagolskih valencijskih okvira obuhvaća opis i popis broja i vrste dopuna kojima određen glagol otvara mjesto u rečenici, odnosno kojima glagol upravlja na sintaktičkome i morfološkome planu. Nakon teorijskoga razmatranja karakteristika valencije i prijelaznosti te pregleda hrvatskih jezičnih resursa koji se na njima temelje, u ovome se radu predstavlja računalni opis glagolskih okvira koji počiva na stvarnoj jezičnoj uporabi. Kao izvor korpusnih podataka odabrana je ovisnosna banka stabala SETimes-HR, trenutačno najveći korpus za hrvatski jezik obilježen na sintaktičkoj razini, koji je označen i mrežno dostupan u sklopu projekta Universal Dependencies (https://universaldependencies.org/hr/). Na primjeru dvoprijelaznih glagola razrađen je model izvlačenja valencijskih okvira uz pomoć paketa pyconll u programskome jeziku Python. Utvrđene su pogreške kod označivanja i poteškoće pri računalnome modeliranju ovisnosne strukture glagola te su predložena rješenja za njih. Kvantitativna analiza dobivenih podataka pokazala je da su dvoprijelazni okviri rijetki u korpusu, ali da se dvoprijelazni glagoli često ostvaruju u drugim valencijskim okvirima. Naposljetku je uveden koncept sintaktičkih n-grama i kolokacija, kojima se mogu proširiti mogućnosti izvlačenja okvira i poboljšati točnost modela. Predstavljena metoda pokazala se korisnom i primjenjivom u budućim istraživanjima valencije i glagolske sintakse te razvoju jezičnih tehnologija za hrvatski jezik.
The goal of the work is to enhance existing financial market forecasting frameworks by including an additional factor–in this example, a collection of carefully chosen tweets—into a long-short repetitive neural channel. In order to produce attributes for such a forecast, this research used a unique attitude analysis approach that combined psychological labelling and a valence rating that represented the strength of the sentiment. Both lexicons produced extra properties such 2-level polarization, 3-level polarization, gross reactivity, as well as total valence. The emotional polarity explicitly marked into the database contrasted well with outcomes of the innovative lexicon approach. Plotting the outcomes of each of these concepts against actual market rates of the equities examined has been the concluding step in this analysis. Root Mean Square Error (RMSE), preciseness, as well as Mean Absolute Percentage Error (MAPE) were used to evaluate the results. Across most instances of market forecasting, attaching an additional factor has been proven to reduce the RMSE and increase the precision of forecasts over lengthy sequences.
International audience
We examined the effects of an informative pitch and multisensory contexts as potential factors influencing individuals’ experience of tofu with soy sauce and the amount consumed outside the lab. Two hundred and sixteen participants watched one of two pitches (promoting either vegetarian diets or exercise) and were guided into one of three multisensory contexts (‘sustainable’, ‘meat’, or ‘neutral’ theme). Participants rated the aroma and appearance of soy sauce and the taste of tofu dipped in it using the intuitive ‘one touch’ EmojiGrid valence and arousal measuring tool. Our results showed that the ‘meat’ context increased arousal ratings for soy sauce and the tendency to consume more tofu relative to the other contexts. Pitch did not influence affective ratings or amounts consumed. We conclude that the multisensory context has the potential to positively affect peoples’ choices and perceptions of plant-based and sustainable food and promote its consumption.
Background Reducing the amount of contrast agent needed for contrast-enhanced breast MRI is desirable. Purpose To investigate if generative adversarial networks (GANs) can recover contrast-enhanced breast MRI scans from unenhanced images and virtual low-contrast-enhanced images. Materials and Methods In this retrospective study of breast MRI performed from January 2010 to December 2019, simulated low-contrast images were produced by adding virtual noise to the existing contrast-enhanced images. GANs were then trained to recover the contrast-enhanced images from the simulated low-contrast images (approach A) or from the unenhanced T1- and T2-weighted images (approach B). Two experienced radiologists were tasked with distinguishing between real and synthesized contrast-enhanced images using both approaches. Image appearance and conspicuity of enhancing lesions on the real versus synthesized contrast-enhanced images were independently compared and rated on a five-point Likert scale. P values were calculated by using bootstrapping. Results A total of 9751 breast MRI examinations from 5086 patients (mean age, 56 years ± 10 [SD]) were included. Readers who were blinded to the nature of the images could not distinguish real from synthetic contrast-enhanced images (average accuracy of differentiation: approach A, 52 of 100; approach B, 61 of 100). The test set included images with and without enhancing lesions (29 enhancing masses and 21 nonmass enhancement; 50 total). When readers who were not blinded compared the appearance of the real versus synthetic contrast-enhanced images side by side, approach A image ratings were significantly higher than those of approach B (mean rating, 4.6 ± 0.1 vs 3.0 ± 0.2; P <.001), with the noninferiority margin met by synthetic images from approach A (P <.001) but not B (P >.99). Conclusion Generative adversarial networks may be useful to enable breast MRI with reduced contrast agent dose. © RSNA, 2023 Supplemental material is available for this article. See also the editorial by Bahl in this issue.
This article reports data from a qualitative study conducted within the elitist English-medium schools in three cities of Pakistan to claim that theory building, and English teaching practices are still modeled on monolingual biases inherent in the orthodox notions of linguistic purism and Anglo-normative traditions of the 1990s. Orthodoxy is manifest in schools’ strong emphasis on direct method and communicative language teaching, and essentialist language management where English is enforced as the sole legitimate language of academic and non-academic transactions. Native language use is attached a stigma and viewed as deficiency in the English language. Such Anglo-normative culture invokes guilty conscience in teachers and students as they disapprove native languages even when they know that they can serve as useful meditating tools for effective content transfer. Crucially, amid English gatekeepers’ celebratory attitudes toward English as the sole legitimate language, some teachers still demonstrate deviation from the pervasive linguistic norms. Showing reflexivity and agency, these teachers ideologically oppose the official policy. They discursively generate spaces for negotiating and appropriating the official policies, which signifies that language policy can be a complex multi-layered process rather than a linear, discrete, or top-down phenomenon. Teachers’ agency is interpreted as ‘decolonial performativity’.
This study aimed to examine whether Japanese participants condition spoken words' meanings to written pseudowords. In Survey 1, we selected spoken words associated with negative (α =.91) and positive (α =.79) features for Experiment 1 and passive (α =.90) and active (α =.80) features for Experiment 2. In Experiment 1, participants evaluated four written pseudowords' emotional valence using a 7-point semantic differential scale (1: negative; 7: positive) before and after conditioning spoken words with negative, neutral, or positive features to each pseudoword. In the conditioning phase, participants read each pseudoword, listened to a spoken word, and verbally repeated each spoken word. The results showed that a pseudoword was conditioned to spoken words with positive and negative features. In Experiment 2, participants evaluated four pseudowords' activeness using a 7-point semantic differential scale (1: passive; 7: active) before and after conditioning spoken words of passive, neutral, and active features to each written pseudoword. In the conditioning phase, the participants read each written pseudoword, listened to a spoken word, and repeated the spoken word. The results showed that the activeness evaluations were more increased for pseudowords conditioned to spoken words of active and neutral features after conditioning than before conditioning but were unchanged for a pseudoword conditioned to those with passive features before and after conditioning. Additonally, Survey 2's results showed that although the positiveness and activeness responses of the words used in Experiments 1 and 2 were controlled well, the lack of significant differences among positiveness responses of words may influence the evaluative conditioning in Experiment 2. That is, when participants condition passive (low arousal) words' activeness (arousal) ratings to those of pseudowords, words' positiveness (valence) ratings would be important in the evaluative conditioning. Our findings suggest that participants can condition spoken word meanings of preference and activeness to those of written pseudowords. It also indicates that linguistically evaluative conditioning's effects are robust in a non-alphabetic language.
The present study aimed to establish and develop an online de novo conditioning paradigm for the measurement of conditioned disgust responses. We further explored the effects of explicit instructions about the CS-UCS contingency on extinction learning and retrieval of conditioned disgust responses. The study included a sample of 115 healthy participants. Geometric figures served as conditioned stimuli (CS) and disgust-evoking pictures as unconditioned stimuli (UCS). During disgust conditioning, the CS+ was paired with the UCS (66% reinforcement) and the CS- remained unpaired; during extinction and retrieval, no UCS was presented. Half of the participants (n = 54) received instructions prior to the disgust extinction stating that the UCS will not be presented anymore. 1-2 days or 7-8 days later participants performed a retrieval test. CS-UCS contingency, disgust and valence ratings were used as dependent measures. Successful acquisition of conditioned disgust response was observed on the level of CS-UCS contingency, disgust and valence ratings. While some decline in valence and disgust ratings during the extinction stage was observed, contingency instructions did not significantly affect extinction performance. Retrieval one week later revealed that contingency instructions increased the discrimination of the CSs. Extinction of conditioned disgust responses is not affected by explicit knowledge of the CS-UCS contingencies. However, contingency instructions prior to extinction seem to have a detrimental effect on long-term extinction retrieval.
The beginnings of words are, in some informal sense, special. This intuition is widely shared, for example, when playing word games. Less apparent is whether the intuition is substantiated empirically and what the underlying organizational principle(s) might be. Here, we answer this seemingly simple question in a quantitatively clear way. Based on arguments about the interplay between lexical storage and speech processing, we examine whether the distribution of information among different speech sounds of words is governed by a critical computational unit for online speech perception and production: syllables. By analyzing lexical databases of twelve languages, we demonstrate that there is a compelling asymmetry between syllable beginnings (onsets) versus ends (codas) in their involvement in distinguishing words stored in the lexicon. In particular, we show that the functional advantage of syllable onset reflects an asymmetrical distribution of lexical informativeness within the syllable unit but not an effect of a global decay of informativeness from the beginning to the end of a word. The converging finding across languages from a range of typological families supports the conjecture that the syllable unit, while being a critical primitive for both speech perception and production, is also a key organizational constraint for lexical storage.
We studied the effect of cutaneous cold stimulus on the perceptual rating of musical chords. Despite the shown influence of music and tactile stimuli on human psychological evaluation, the effect of a cold stimulus on sound perception remains underexplored. We examined the effect of a cold stimulus on four psychological measures (frisson, arousal, pleasantness, and valence) as participants listened to two-note chords (consonance and dissonance). The cold-stimulus condition involved an experimenter touching the back of the participant's neck with a cooling device while listening to the sounds, while the control condition used a cooling device with the power off. For the frisson and arousal measures, the main effect of the stimulus condition was significant, showing that the cold stimulus increased the frisson and arousal measures. For the pleasantness and valence measures, there was a significant main effect of two-note chords, showing that a consonance was perceived as more pleasant than a dissonance; however, there was no significant main effect of stimulus condition, showing that the cold stimulus did not affect pleasantness and valence ratings. The results showed that a cold stimulus could bias frisson and arousal without affecting pleasantness and valence ratings when listening to musical sound.
Languages are known to describe the world in diverse ways. Across lexicons, diversity is pervasive, appearing through phenomena such as lexical gaps and untranslatability. However, in computational resources, such as multilingual lexical databases, diversity is hardly ever represented. In this paper, we introduce a method to enrich computational lexicons with content relating to linguistic diversity. The method is verified through two large-scale case studies on kinship terminology, a domain known to be diverse across languages and cultures: one case study deals with seven Arabic dialects, while the other one with three Indonesian languages. Our results, made available as browseable and downloadable computational resources, extend prior linguistics research on kinship terminology, and provide insight into the extent of diversity even within linguistically and culturally close communities.
INTRODUCTION: Social anxiety disorder (SAD) is characterized by abnormal processing of performance-related social stimuli. Previous studies have shown altered emotional experiences and activations of different sub-regions of the striatum during processing of social stimuli in patients with SAD. However, whether and to what extent social comparisons affect behavioural and neural responses to feedback stimuli in patients with SAD is unknown. MATERIALS AND METHODS: To address this issue, emotional ratings and functional magnetic resonance imaging (fMRI) responses were assessed while patients suffering from SAD and healthy controls (HC) were required to perform a choice task and received performance feedback (correct, incorrect, non-informative) that varied in relation to the performance of fictitious other participants (a few, half, or most of others had the same outcome). RESULTS: Across all performance feedback conditions, fMRI analyses revealed reduced activations in bilateral putamen when feedback was assumed to be received by only a few compared to half of the other participants in patients with SAD. Nevertheless, analysis of rating data showed a similar modulation of valence and arousal ratings in patients with SAD and HC depending on social comparison-related feedback. CONCLUSIONS: This suggests altered neural processing of performance feedback depending on social comparisons in patients with SAD.
Social support offers protection from depression and anxiety, possibly through its beneficial effects upon cognitive reappraisal. The present study evaluates potential mechanisms of social support, utilizing a reappraisal task in 121 undergraduates elevated on neuroticism. Participants were instructed to reinterpret stressful images with (Social Condition) and without (Solo Condition) the reminder of a social support figure. Aversiveness, negative affect, and positive affect ratings, as well as written reappraisal responses, were collected trial-by-trial. Results indicated that participants reported lower aversiveness and negative affect and higher positive affect when reinterpreting images in the Social Condition compared to the Solo Condition. Analyses of adherence ratings of written reappraisals revealed that participants generated reinterpretations more in the Social Condition than in the Solo Condition. Exploratory mediation analyses indicated that Condition was indirectly associated with reappraisal efficacy as measured by aversiveness and affect ratings through reappraisal adherence. Results suggest that cognitive reappraisal with the influence of social support may be more effective than cognitive reappraisal without such social influence, and thus may be a suitable target for interventions for depression and anxiety.
Oscillatory synchronization in the theta-frequency band was found to play a causal role in binding information of different modalities in declarative memory. Moreover, there is first evidence from a laboratory study that theta-synchronized (vs. asynchronized) multimodal input in a classical fear conditioning paradigm resulted in better discrimination of a threat-associated stimulus when compared to perceptually similar stimuli never associated with the aversive unconditioned stimulus (US). Effects manifested in affective ratings and ratings of contingency knowledge. However, theta-specificity was not addressed so far. Thus, in the present pre-registered web-based fear conditioning study, we compared synchronized (vs. asynchronized) input in a theta-frequency band vs. the same synchronization manipulation in a delta frequency. Based on our previous laboratory design, five visual gratings of different orientations (25°, 35°, 45°, 55°, 65°) served as conditioned stimuli (CS) with only one (CS+) paired with the auditory aversive US. Both CS and US were luminance or amplitude modulated, respectively, in a theta (4 Hz) or delta (1.7 Hz) frequency. In both frequencies, CS-US pairings were presented either in-phase (0° phase lag) or out-of-phase (90°, 180°, 270°), resulting in four independent groups (each group N = 40). Phase synchronization augmented the discrimination of CSs in CS-US contingency knowledge but did not affect valence and arousal ratings. Interestingly, this effect occurred independent of frequency. In sum, the current study proves the ability to successfully conduct complex generalization fear conditioning in an online setting. Based on this prerequisite, our data supports a causal role of phase synchronization in the declarative CS-US associations for low frequencies rather than in the specific theta-frequency band.
Introduction Deep brain stimulation (DBS) studies in Parkinson's Disease (PD) targeting the subthalamic nucleus (STN) have characterized its spectral properties across cognitive processes. In emotional evaluation tasks, specific alpha frequency (8–12 Hz) event-related de-synchronization (ERD) (reduced power) has been demonstrated. The time-locked stimulation of STN relative to stimuli onset has shown subjective positive valence shifts with 10 Hz but not with 130 Hz. However, neurophysiological effects of stimulation on power modulation have not been investigated. We aim to investigate effects of acute stimulation of the right STN on concurrent power modulation in the contralateral STN and frontal scalp EEG. From our previous study, we had a strong a priori hypothesis that negative imagery without stimulation would be associated with alpha ERD; negative imagery with 130 Hz stimulation would be also associated with alpha ERD given the lack of its effect on subjective valence ratings; negative imagery with 10 Hz stimulation was to be associated with enhanced alpha power given the shift in behavioral valence ratings. Methods Twenty-four subjects with STN DBS underwent emotional picture-viewing tasks comprising neutral and negative pictures. In a subset of these subjects, the negative images were associated with time-locked acute stimulation at either 10 or 130 Hz. Power of signals was estimated relative to the baseline and subjected to non-parametric statistical testing. Results As hypothesized, in 130 Hz stimulation condition, we show a decrease in alpha power to negative vs. neutral images irrespective of stimulation. In contrast, this alpha power decrease was no longer evident in the negative 10 Hz stimulation condition consistent with a predicted increase in alpha power. Greater beta power in the 10 Hz stimulation condition along with correlations between beta power across the 10 Hz stimulation and unstimulated conditions suggest physiological and cognitive generalization effects. Conclusion Acute alpha-specific frequency stimulation presumably was associated with a loss of this expected decrease or desynchronization in alpha power to negative images suggesting the capacity to facilitate the synchronization of alpha and enhance power. Acute time-locked stimulation has the potential to provide causal insights into the spectral frequencies and temporal dynamics of emotional processing.
Background When receiving mismatching nonverbal and verbal signals, most people tend to base their judgment regarding the current emotional state of others primarily on nonverbal information. However, individuals with high-functioning autism (HFA) have been described as having difficulties interpreting nonverbal signals. Recognizing emotional states correctly is highly important for successful social interaction. Alterations in perception of nonverbal emotional cues presumably contribute to misunderstanding and impairments in social interactions. Methods To evaluate autism-specific differences in the relative impact of nonverbal and verbal cues, 18 adults with HFA (14 male and four female subjects, mean age 36.7 years (SD 11.4) and 18 age, gender and IQ-matched typically developed controls [14 m/4 f, mean age 36.4 years (SD 12.2)] rated the emotional state of speakers in video sequences with partly mismatching emotional signals. Standardized linear regression coefficients were calculated as a measure of the reliance on the nonverbal and verbal components of the videos for each participant. Regression coefficients were then compared between groups to test the hypothesis that autistic adults base their social evaluations less strongly on nonverbal information. Further exploratory analyses were performed for differences in valence ratings and response times. Results Compared to the typically developed control group, nonverbal cue reliance was reduced in adults with high-functioning autism [ t (23.14) = −2.44, p = 0.01 (one-sided)]. Furthermore, the exploratory analyses showed a tendency to avoid extreme answers in the HFA group, observable via less positive as well as less negative valence ratings in response to emotional expressions of increasingly strong valence. In addition, response time was generally longer in HFA compared to the control group [ F (1, 33) = 10.65, p = 0.004]. Conclusion These findings suggest reduced impact of nonverbal cues and longer processing times in the analysis of multimodal emotional information, which may be associated with a subjectively lower relevance of this information and/or more processing difficulties for people with HFA. The less extreme answering tendency may indicate a lower sensitivity for nonverbal valence expression in HFA or result from a tendency to avoid incorrect answers when confronted with greater uncertainty in interpreting emotional states.
Assessing integrated reading-into-writing task performances is known to be challenging, and analytic rating scales have been found to better facilitate the scoring of these performances than other common types of rating scales. However, little is known about how specific operationalizations of the reading-into-writing construct in analytic rating scales may affect rating quality, and by extension score inferences and uses. Using two different analytic rating scales as proxies for two approaches to reading-into-writing construct operationalization, this study investigated the extent to which these approaches affect rating reliability and consistency. Twenty raters rated a set of reading-into-writing performances twice, each time using a different analytic rating scale, and completed post-rating questionnaires. The findings resulting from our convergent explanatory mixed-method research design show that both analytic rating scales functioned well, further supporting the use of analytic rating scales for scoring reading-into-writing. Raters reported that either type of analytic rating scale prompted them to attend to the reading-related aspects of reading-into-writing, although rating these aspects remained more challenging than judging writing-related aspects. The two scales differed, however, in the extent to which they led raters to uniform interpretations of performance difficulty levels. This study has implications for reading-into-writing scale design and rater training.
We present CamelParser2.0, an open-source Python-based Arabic dependency parser targeting two popular Arabic dependency formalisms, the Columbia Arabic Treebank (CATiB), and Universal Dependencies (UD). The CamelParser2.0 pipeline handles the processing of raw text and produces tokenization, part-of-speech and rich morphological features. As part of developing CamelParser2.0, we explore many system design hyper-parameters, such as parsing model architecture and pretrained language model selection, achieving new state-of-the-art performance across diverse Arabic genres under gold and predicted tokenization settings.
Psycholinguistic research on isiXhosa and related Bantu languages is scarce. For research on lexical processing in particular, a prerequisite is data on lexical properties that impact word recognition, such as word frequency and neighbourhood density. This paper describes the derivation of these and related lexical statistics from a newly created 4.8-million-word isiXhosa corpus. It then reviews the potential applications of such a lexical database for research on language acquisition, language development, and language processing. The paper closes with recommendations for further work in this domain.