Papers reviewed and determined not to be word norm studies. Use the flag icon to report errors or suggest re-inclusion.
18265 papers
People experience the same event but do not feel the same way. Such individual differences in emotion response are believed to be far greater than those in any other mental functions. Thus, to understand what makes people individuals, it is important to identify the systematic structures of individual differences in emotion response and elucidate how such structures relate to what aspects of psychological characteristics. Reflecting this importance, many studies have attempted to relate emotions to psychological characteristics such as personality traits, psychosocial states, and pathological symptoms across individuals. However, systematic and global structures that govern the across-individual covariation between the domain of emotion responses and that of psychological characteristics have been rarely explored previously, which limits our understanding of the relationship between individual differences in emotion response and psychological characteristics. To overcome this limitation, we acquired high-dimensional data sets in both emotion-response (8 measures) and psychological-characteristic (68 measures) domains from the same pool of individuals (86 undergraduate or graduate students) and carried out the canonical correlation analysis in conjunction with the principal component analysis on those data sets. For each participant, the emotion-response measures were quantified by regressing affective-rating responses to visual narrative stimuli onto the across-participant average responses to those stimuli, while the psychological-characteristic measures were acquired from 19 different psychometric questionnaires grounded in personality, psychosocial-factor, and clinical-problem taxonomies. We found a single robust mode of population covariation, particularly between the 'accuracy' and 'sensitivity' measures of arousal responses in the emotion domain and many 'psychosocial' measures in the psychological-characteristics domain. This mode of covariation suggests that individuals characterized with positive social assets tend to show polarized arousal responses to life events.
Music is capable of conveying many emotions. The level and type of emotion of the music perceived by a listener, however, is highly subjective. In this study, we present the Music Emotion Recognition with Profile information dataset (MERP). This database was collected through Amazon Mechanical Turk (MTurk) and features dynamical valence and arousal ratings of 54 selected full-length songs. The dataset contains music features, as well as user profile information of the annotators. The songs were selected from the Free Music Archive using an innovative method (a Triple Neural Network with the OpenSmile toolkit) to identify 50 songs with the most distinctive emotions. Specifically, the songs were chosen to fully cover the four quadrants of the valence arousal space. Four additional songs were selected from DEAM to act as a benchmark in this study and filter out low quality ratings. A total of 277 participants participated in annotating the dataset, and their demographic information, listening preferences, and musical background were recorded. We offer an extensive analysis of the resulting dataset, together with a baseline emotion prediction model based on a fully connected model and an LSTM model, for our newly proposed MERP dataset.
Universal Dependencies is an international community project and a collection of morphosyntactically annotated data sets (“treebanks”) for more than 100 languages. The collection is an invaluable resource for various linguistic studies, ranging from grammatical constructions within one language to language typology, documentation of endangered languages, and historical evolution of language. In the tutorial, I will first quickly show the main principles of UD, then I will present the actual d...
Most computational models of dependency syntax consist of distributions over spanning trees. However, the majority of dependency treebanks require that every valid dependency tree has a single edge coming out of the ROOT node, a constraint that is not part of the definition of spanning trees. For this reason all standard inference algorithms for spanning trees are sub-optimal for inference over dependency trees.Zmigrod et al (2021) proposed algorithms for sampling with and without replacement from the dependency tree distribution that incorporate the single-root constraint. In this paper we show that their fastest algorithm for sampling with replacement, Wilson-RC, is in fact producing biased samples and we provide two alternatives that are unbiased. Additionally, we propose two algorithms (one incremental, one parallel) that reduce the asymptotic runtime of algorithm for sampling k trees without replacement to O(kn^3). These algorithms are both asymptotically and practically more efficient.
Abstract When designing a sign language dictionary or lexical database, a criterion that needs to be decided on is how to organize handshapes in the search-by-sign interface. Although notation systems do exist for indexing and searching sign entries, dictionaries do not often use them. This study investigates how to classify and order images of handshapes without relying on the alphabetical order of handshape names or phonological parameters. For this purpose, a cluster analysis using three variables was applied, resulting in groups based on visual similarities between handshapes. The objects in the resulting clusters were reordered based on similarity and shape gradation principles, aiming for an optimal organization with regard to Prägnanz. Ultimately, as we expect native signers to rely on the visual features of signs, the taxonomy proposed in this study makes it possible to display handshapes in a way that requires fewer metalinguistic skills to support signers’ search actions.
As input representation for each sub-word, the original BERT architecture proposes the sum of the sub-word embedding,<br/>position embedding and a segment embedding. Sub-word and position embeddings are well-known and studied, and<br/>encode lexical information and word position, respectively. In contrast, segment embeddings are less known and have so<br/>far received no attention. The key idea of segment embeddings is to encode to which of the two sentences (segments)<br/>a word belongs—the intuition is to inform the model about the separation of sentences for the next sentence prediction<br/>pre-training task. However, little is known on whether the choice of segment impacts downstream prediction performance.<br/>In this work, we try to fill this gap and empirically study the impact of alternating the segment embedding during inference<br/>time for a variety of pre-trained embeddings and target tasks. We hypothesize that for single-sentence prediction tasks<br/>performance is not affected—neither in mono- nor multilingual setups—while it matters when changing the segment IDs<br/>in paired-sentence tasks. To our surprise, this is not the case. Although for classification tasks and monolingual BERT<br/>models no large differences are observed, particularly word-level multilingual prediction tasks are heavily impacted. For<br/>low-resource syntactic tasks, we observe impacts of segment embedding and multilingual BERT choice. We find that<br/>the default setting for the most used multilingual BERT model underperforms heavily, and a simple swap of the segment<br/>embeddings yields an average improvement of 2.5 points absolute LAS score for dependency parsing over 9 different treebanks.
This study examines the approaches to defining the notion of strangeness from the standpoints of philosophy, intercultural communication, and translation studies. Depending on intensity of axiological characteristics by such criteria as space, affiliation, trust, and compliance with linguistic norms, strangeness is verbalized by expanding the "own:: other – foreign" opposition. On the textual level, "otherness" and "foreignness" are characterized by borderline, subjective perception of the recipient; meanwhile, it is noted that the translation decision on overcoming strangeness is made directly by the translator as the initial recipient. Interactions with representatives of other countries and cultures are an active part of the academic discourse, defined by the authors as communication in the high school space and its educational, research, and sociocultural areas. This paper focuses on the international presentational discourse of high school, which implements its phatic tasks in the informative texts posted on an educational institutions website in the original and translated versions. As participants of the presentational discourse are inclined to have a dialog, strangeness tends to be perceived as a form of otherness. Its explication in translated texts is performed by full and functionally adequate translation of the units, such as realia, forms of address, neologisms, proper nouns, which communicate unique information and reflect linguacultural peculiarities.
PURPOSE: Emotional processing allows us to predict our own and others' behavior, communicate our wants and needs, and understand those of others. Thus, deficits in emotional processing can negatively impact one's quality of life. While changes in emotional processing across several domains (e.g., prosody, faces) in Parkinson's disease (PD) are widely accepted, there is a dearth of literature, with equivocal results, regarding how emotional language processing is affected by PD. This study investigated emotional sentence processing in this population. METHOD: Eighteen persons with PD and 22 healthy adults (HAs) completed a language task in which they rated sentences on their pleasantness (valence), and a battery of cognitive tasks and mood measures that were examined as factors influencing performance. As an interaction between emotionality and concreteness during processing has been indicated in prior research, concreteness of sentence stimuli was also manipulated. RESULTS: Individuals with PD rated negatively valenced sentences as less negative and positively-valenced sentences as less positive than HAs. The PD group also demonstrated a reduced overall range of valence rating scores. Sentence concreteness did not influence ratings. Results for positive sentences could be explained by individual differences in working memory (WM), whereas individual differences in WM, depression, and group explained differences in ratings to negative sentences. CONCLUSIONS: Our study provides one of few accounts of emotional language processing deficits in PD, particularly beyond the word level. Individuals with PD may experience difficulty perceiving and assessing the intensity of the emotional content of language, and deficits may disproportionately impact processing of sentences about negative situations. SUPPLEMENTAL MATERIAL: https://doi.org/10.23641/asha.21313713.
OBJECTIVE: Facial affect recognition is associated with neuropsychological status and psychiatric diseases. We hypothesized that facial affect recognition is associated with psychological status and perception of other affects. METHODS: A total of 80 images depicting facial affect, including 20 Neutral, 20 Angry, 20 Fear, and 20 Sad, were screened for use in our research. A total of 100 healthy individuals were asked to rate these images using a 10-point Likert scale and complete psychological scales assessing the emotional statuses and cognitive functions. RESULTS: The participants' emotional state of aggression, attention, and impulsivity may have been associated with their interpretation of the Angry facial expressions. The participants often rated the Angry facial expressions as Fear. The participants rated Fear images as Angry or Sad. In response to a Sad facial expression, the participants reported psychological statuses of attention and impulsivity which were associated with the facial expression rating. The participants rated the Sad expression as Angry or Fear. CONCLUSION: The psychological statuses of the participants were significantly correlated with their interpretation of facial affects. In particular, a psychological state of attention was often correlated with incorrect affect ratings. Attention and impulsivity could affect the rating of the sad facial expressions.
The article attempts to show how the theory of Frame Semantics and the resources of the lexical database FrameNet can be used for teaching/learning terminology of specialised domains. The article discusses the principles of Frame Semantics and presents a use case of application of the frame-based methodology for developing classification of terminology of the selected financial subdomain for learning/teaching purposes. The use case focuses on terms denoting concepts that compose ‘CAUSE-RISK’ frame which was developed on the basis of several related frames in the FrameNet database. The stages of the use case and its outcomes are described in detail and the benefits of application of the methodology for learning/teaching specialised vocabulary are provided. Hopefully, the provided insights will give ideas to teachers of foreign languages for specific purposes and help to develop effective terminology teaching/learning techniques.
The field of Labovian quantitative sociolinguistics, also referred to as sociolinguistic studies of Language Variation and Change (hereafter LVC), to the extent it uses the term at all, understands linguistic norms as norms of production and perception and, crucially, evaluation of linguistic performance. These norms and the behaviors they constrain have traditionally played a central role in circumscribing speech communities, the bounded sets of speakers that feature in sociolinguistic studies. In this conception, some combination of norms will be shared within, but not outside, a speech community, at whatever scale we are considering. The establishment of these boundaries is then a matter of empirical investigation. In this case, scale can matter, because some patterns of variable production will be nation-wide (or worldwide), while others will be more local. The indeterminate and perception-driven status of the speech community comes out in Eckert’s definition of the speech community as “an aggregate of people who are committed to talking the same, in the face of the fact that they don’t quite” (2018: 165). Speakers are hereby participating in a belief (or ideology, or tacit understanding) that all members speak the same way, and in that sense are conforming to norms that delineate the speech community, although performance, measured quantitatively and statistically, will not be identical for all speakers. Sociolinguistic norms (and deviance from them) also carry indexical and ideological baggage, as Fabricius and Mortensen (2013) and Mortensen (2014) demonstrate. The concept of ‘construct resource’ discussed in those papers will be revisited here in a discussion of to what extent we can understand processes of sociolinguistic change as norm–transformation. The paper considers the example of the trajectory of change in prevocalic /r/ in Received Pronunciation, the elite sociolect of the United Kingdom.
Abstract In many respects Latin poetry deviates from the linguistic norms of prose; such deviations are often attributed to non-linguistic factors, with the result that the language of Latin poetry is more often studied by literary scholars than by linguists. A frequent poetic phenomenon are discontinuous adjective-noun phrases, which tend to be explained by appeal to various artistic considerations. The present contribution argues that syntactic discontinuity in poetry is not arbitrary, but is subject to linguistic constraints, which are largely the same as in prose. By scrutinising 120 non-discontinuous adjective-noun phrases in the Ciris (out of a total of 531 adjective-noun phrases), it is demonstrated that the overwhelming majority of such phrases fall into a number of specific syntactic and semantic categories, while outside these categories discontinuity appears to be the rule.
OBJECTIVE: Examine cochlear implant (CI) users' ability to identify safety-relevant environmental sounds, imperative for safety, independence, and personal well-being. METHODS: Twenty-one experienced adult CI users completed an Environmental Sound Identification (ESI) test consisting of 42 common environmental sounds, 28 of which were relevant to personal safety, along with 14 control sounds. Prior to sound identification, participants were shown sound names and asked to rate the familiarity and, separately, relevance to safety of each corresponding sound on a 1-5 scale. RESULTS: Overall ESI accuracy was 57% correct for the safety-relevant sounds and 55% correct for control sounds. Participants rated safety-relevant sounds as more important to safety and more familiar than the non-safety sounds. ESI accuracy significantly correlated with familiarity ratings. CONCLUSION: The present findings suggest mediocre ESI accuracy in postlingual adult CI users for safety-relevant and other environmental sounds. Deficits in the identification of these sounds may put CI listeners at increased risk of accidents or injuries and may require a specific rehabilitation program to improve CI outcomes. LEVEL OF EVIDENCE: 4 Laryngoscope, 133:2388-2393, 2023.
Prior research has discussed and illustrated the need to consider linguistic norms at the community level when studying taboo (hateful/offensive/toxic etc.) language. However, a methodology for doing so, that is firmly founded on community language norms is still largely absent. This can lead both to biases in taboo text classification and limitations in our understanding of the causes of bias. We propose a method to study bias in taboo classification and annotation where a community perspective is front and center. This is accomplished by using special classifiers tuned for each community's language. In essence, these classifiers represent community level language norms. We use these to study bias and find, for example, biases are largest against African Americans (7/10 datasets and all 3 classifiers examined). In contrast to previous papers we also study other communities and find, for example, strong biases against South Asians. In a small scale user study we illustrate our key idea which is that common utterances, i.e., those with high alignment scores with a community (community classifier confidence scores) are unlikely to be regarded taboo. Annotators who are community members contradict taboo classification decisions and annotations in a majority of instances. This paper is a significant step toward reducing false positive taboo decisions that over time harm minority communities.
Developing semantic hierarchies from user-created hashtags in social media can provide useful organizational structure to large volumes of data. However, construction of these hierarchies is difficult using established ontologies (e.g. WordNet [C. Fellbaum (ed.), WordNet: An Electronic Lexical Database (MIT Press, Cambridge, MA, 1998)]) due to the differences in the semantic and pragmatic use of words versus hashtags in social media. While alternative construction methods based on hashtag frequency are relatively straightforward, these methods can be susceptible to the dynamic nature of social media, such as hashtags with brief surges in popularity. We drew inspiration from the ecologically based Shannon Diversity Index (SDI) [J. L. Wilhm, Use of biomass units in Shannon’s formula, Ecology 49(1) (1968) 153–156] to create a more representative and resilient method of semantic hierarchy construction that relies upon network-based community detection and a novel, entropy-based ensemble diversity index (EDI) score. The EDI quantifies the contextual diversity of each hashtag, resulting in thousands of semantically related groups of hashtags organized along a general-to-specific spectrum. Through an application of EDI to social media data (Twitter and Parler) and a comparison of our results to prior approaches, we demonstrate our method’s ability to create semantically consistent hierarchies that can be flexibly applied and adapted to a range of use cases.
Abstract Gender-inclusive language, both binary and non-binary, advocates for wider visibility of non-dominant genders. However, in the Spanish context, this language, especially the binary variant, has been received with much opposition led by the institution establishing linguistic norms. This paper addresses the possibility of using gender-inclusive language as a political practice in translation and explores how entertainment genres, particularly comics, and information on the social goals of heterodox linguistic practices may be conducive to relaxing the adherence to dominant doxas. By priming responses based on two gender-inclusive (one binary and one non-binary) translations of the comic Morgane by Kansara and Fert, differences between four focus groups were analyzed. Two of them received a briefing session detailing the social agenda of gender-inclusive language, and two were offered no such sessions before reading and discussing the translations. Results showed how the briefing sessions relaxed adherence to the dominant doxa even though the social purpose of gender-inclusive language was widely questioned. Further, results evinced that binary translation was more strongly opposed than the non-binary variant. It is argued that translation can be used as a tool for political and social change and become instrumental in advocating for equal opportunities for all genders.
Estimating the semantic similarity between short texts plays an increasingly prominent role in many fields related to text mining and natural language processing applications, especially with the large increase in the volume of textual data that is produced daily. Traditional approaches for calculating the degree of similarity between two texts, based on the words they share, do not perform well with short texts because two similar texts may be written in different terms by employing synonyms. As a result, short texts should be semantically compared. In this paper, a semantic similarity measurement method between texts is presented which combines knowledge-based and corpus-based semantic information to build a semantic network that represents the relationship between the compared texts and extracts the degree of similarity between them. Representing a text as a semantic network is the best knowledge representation that comes close to the human mind's understanding of the texts, where the semantic network reflects the sentence's semantic, syntactical, and structural knowledge. The network representation is a visual representation of knowledge objects, their qualities, and their relationships. WordNet lexical database has been used as a knowledge-based source while the GloVe pre-trained word embedding vectors have been used as a corpus-based source. The proposed method was tested using three different datasets, DSCS, SICK, and MOHLER datasets. A good result has been obtained in terms of RMSE and MAE.
Abstract Examining peoples’ affect and emotions over time and their effects on peoples’ behavior are ongoing endeavors in human-computer-interaction (HCI) research. This paper reports an experiment in which participants watched either positive or negative film clips on a tablet PC to enter a positive or negative affective state. Successively, they accomplished four basic system interaction tasks like changing fonts of an app on the same device. Results show that, in line with previous studies, peoples’ general valence ratings quickly reverted to neutral when starting the task accomplishment. At the level of distinct positive emotions, participants’ ratings of hope, joy, and serenity decreased after watching negative film clips. Moreover, amusement, love, and serenity decreased during the interaction with the tablet PC. Amongst the negative emotions, only ratings of sadness increased after watching negative film clips and decreased again after the interaction. Also, participants in the positive film group were slower in executing one of the basic tasks than participants in the negative film group. The findings suggest that only few emotions may be causal for peoples’ ratings of general affect. Results also indicate that negative emotions may help people executing standard tasks, in contrast to positive emotions. Implications for HCI design and research are discussed.
Background & aims: Throughout typical development, children prioritize different perceptual, social, and linguistic cues to learn words. The earliest acquired words are often those that are perceptually salient and highly imageable. Imageability, the ease in which a word evokes a mental image, is a strong predictor for word age of acquisition in typically developing (TD) children, independent of other lexicosemantic features such as word frequency. However, little is known about the effects of imageability in children with autism spectrum disorder (ASD), who tend to have differences in linguistic processing and delayed language acquisition compared to their TD peers. This study explores the extent to which imageability and word frequency are associated with early noun and verb acquisition in children with ASD. Methods: Secondary analyses were conducted on previously collected data of 156 children (78 TD, 78 ASD) matched on sex and parent-reported language level. Total expressive vocabulary, as measured by the MacArthur Bates Communicative Development Inventory (MB-CDI), included 123 words (78 nouns, 45 verbs) that overlapped with previously published imageability ratings and word input frequencies. A two-step hierarchical linear regression was used to examine the relationship between word input frequency, imageability, and total expressive vocabulary. An F-test was then used to assess the unique contribution of imageability on total expressive vocabulary when controlling for word input frequency. Results: In both the TD and ASD groups, imageability uniquely explained a portion of the variance in total expressive vocabulary size, independent of word input frequency. Notably, imageability was significantly associated with noun vocabulary and verb vocabulary size alone, with imageability explaining a greater portion of the variance in total nouns produced than in total verbs produced. Conclusions: Imageability was identified as a significant lexicosemantic feature for describing expressive vocabulary size in children with ASD. Consistent with literature on TD children, children with ASD who have small vocabularies primarily produce words that are highly imageable. Children who are more proficient word learners with larger vocabularies produce words that are less imageable, indicating a potential shift away from reliance on perceptual-based language processing. This was consistent across both noun and verb vocabularies. Implications: Our findings contribute to a growing body of literature describing early word learning in children with ASD and provide a basis for exploring the use of multisensory language learning strategies.
According to the Menzerath-Altmann law, the mean word length is greater in shorter clauses than in longer ones. In Czech, negation is mostly realized by adding the prefix ne- to the beginning of the word, which makes the word longer (and, consequently, it also increases the mean word length in the clause). Therefore, we predict that clauses in which the predicate is in the affirmative form are longer than ones with the negative predicate. We test the hypothesis on a sample of 59 pairs of affirmative and negative forms of the same verb from the Prague Dependency Treebank 3.0.
This article juxtaposes the reflections on the expectations from plain language as declared by the participants of workshops delivered by the Jasnopis team with the status of linguistic research on phenomena such as linguistic awareness, linguistic norm, or attitudes towards language. These arguments serve the purpose of proposing a comprehensive approach to text, which positions the recipient in the centre of works on plain language, which allows for the output of other disciplines of knowledge and requires cooperation between experts and practitioners in a given field, text recipients, and linguists/editors. Such an approach is put forward also by the plain language ISO standard (due for publication), which encourages examination of a text together with its recipients and employment of non-linguistic information design techniques. Keywords: plain language – ISO 24495-1 – comprehensive approach to text.
Perception of passing time can be distorted1. Emotional experiences, particularly arousal, can contract or expand experienced duration, via its interactions with attentional and sensory processing mechanisms2,3. Current models suggest that perceived duration is constructed from accumulation processes4,5 and is encoded from temporally evolving neural dynamics6,7. 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 processing8-15. 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 categorised as short or long16. 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, while 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 towards contraction. Thus, experienced time contracts and expands within each heartbeat – a balance that is disrupted under heightened arousal.
Prior experience represents a prerequisite for memory consolidation across various memory systems. In the context of olfaction, sleep was found to enhance the consolidation of odors in adults but not in typically developing children (TDC), likely due to differences in pre-experience. Interestingly, unmedicated children with attention deficit hyperactivity disorder (ADHD), a neurodevelopmental condition related to dopamine dysfunction, showed lower perceptive thresholds for odors, potentially allowing for more odor experience compared to TDC. We investigated sleep-associated odor memory consolidation in ADHD. Twenty-eight children with ADHD and thirty age-matched TDC participated in an incidental odor recognition task. For the sleep groups (ADHD: n = 14, TDC: n = 15), the encoding of 10 target odorants took place in the evening, and the retention of odorants was tested with 10 target odorants and 10 distractor odorants the next morning. In the wake groups (ADHD: n = 14, TDC: n = 15), the time schedule was reversed. Odor memory consolidation was superior in the ADHD sleep group compared to the TDC sleep and the ADHD wake groups. Intensity and familiarity ratings during encoding were substantially higher in ADHD compared to TDC. Sleep-associated odor memory consolidation in ADHD is superior to TDC. Abundant pre-experience due to lower perceptive thresholds is suggested as a possible explanation. Olfaction might serve as a biomarker in ADHD.
Recently, AWD-LSTM (ASGD Weight-Dropped LSTM) has achieved good result in the language model, and many AWD-LSTM based models have obtained state-of-the-art perplexities. However, in fact, large-scale neural language models have been shown to be prone to overfitting. In AWD-LSTM original paper, the author decided to adopt the way of retraining calling finetune to get a better result. In this paper, we present a simple yet effective parameter rollback mechanism for neural language models. And we introduce the parameter rollback averaged stochastic gradient descent (PR-ASGD), wherein the parameter “step” in ASGD will decrease according to a certain probability. Using this strategy, we achieve better word level perplexities on Penn Treebank: 56.26 based on AWD-LSTM model and 53.57 based on AWD-LSTM-MoS (AWD-LSTM Mixture of Softmaxes) model.
Recurrent Neural Networks (RNNs) have become important tools for tasks such as speech recognition, text generation, or natural language processing. However, their inference may involve up to billions of operations and their large number of parameters leads to large storage size and runtime memory usage. These reasons impede the adoption of these models in real-time, on-the-edge applications. Field-Programmable Gate Arrays (FPGAs) and Application-Specific Integrated Circuits (ASICs) have emerged as promising solutions for the hardware acceleration of these algorithms, thanks to their degree of customization of compute data paths and memory subsystems, which makes them take the maximum advantage from compression techniques for what concerns area, timing, and power consumption. In contrast to the extensive study in compression and quantization for plain feed forward neural networks in the literature, little attention has been paid to reducing the computational resource requirements of RNNs. This work proposes a new effective methodology for the post-training quantization of RNNs. In particular, we focus on the quantization of Long Short-Term Memory (LSTM) RNNs and Gated Recurrent Unit (GRU) RNNs. The proposed quantization strategy is meant to be a detailed guideline toward the design of custom hardware accelerators for LSTM/GRU-based algorithms to be implemented on FPGA or ASIC devices using fixed-point arithmetic only. We applied our methods to LSTM/GRU models pretrained on the IMDb sentiment classification dataset and Penn TreeBank language modelling dataset, thus comparing each quantized model to its floating-point counterpart. The results show the possibility to achieve up to 90% memory footprint reduction in both cases, obtaining less than 1% loss in accuracy and even a slight improvement in the Perplexity per word metric, respectively. The results are presented showing the various trade-offs between memory footprint reduction and accuracy changes, demonstrating the benefits of the proposed methodology even in comparison with other works from the literature.
The emergence of world Englishes (WE) has revolutionized the way scholars and linguists view the English language. Before Kachru introduced the concentric circles model of WE, there were only two varieties of English that were widely accepted in the world, namely, American English (AmE) and British English—both of which fall in the category of Inner Circle Englishes. Despite having its own linguistic norms, PhE is yet to be recognized and accepted in the discipline of language assessment. This chapter argues that language assessment in the Philippines must be WE paradigm-informed. The resistance to include appropriate linguistic norms is one of the issues in testing WE. Despite the fact that the variety of English being used in Philippine classrooms is PhE, the assessment tasks and standards taught are still those of AmE.
Modern Irish is a minority language lacking sufficient computational resources for the task of accurate automatic syntactic parsing of usergenerated content such as tweets. Although language technology for the Irish language has been developing in recent years, these tools tend to perform poorly on user-generated content. As with other languages, the linguistic style observed in Irish tweets differs, in terms of orthography, lexicon, and syntax, from that of standard texts more commonly used for the development of language models and parsers. We release the first Universal Dependencies treebank of Irish tweets, facilitating natural language processing of user-generated content in Irish. In this paper, we explore the differences between Irish tweets and standard Irish text, and the challenges associated with dependency parsing of Irish tweets. We describe our bootstrapping method of treebank development and report on preliminary parsing experiments.
In Japan, the increase in the number of elderly people with dementia due to the aging of society has become a major social problem. This study designed a recreation using an animal robot to increase the amount of communication among elderly people with dementia. Then, we conducted the proposal recreation at a nursing home for elderly people with dementia to measure the effectiveness of the proposed recreation. The effectiveness of the recreation was measured not only by increasing the amount of communication, but also by evaluating the emotions of the elderly people with dementia while they were participating in the recreation. Lawton’s Philadelphia Geriatric Center Affect Rating Scale ("ARS") was used in the evaluation of emotions. The results confirmed that recreation using the animal robot increased the amount of communication among elderly people with dementia, and increased positive emotions among elderly people with dementia.
Obo-Manobo, one of the tribes in Mindanao Philippines, inhabiting some municipalities in the Province of North Cotabato, has ethnic or indigenous language which is not sufficiently analyzed and documented. To account varieties in language and to come up with the glossary of terms, the researcher used the Metalinguistic knowledge from the interviews, focus group discussions, observations and translation of the Swadesh Word List and from the lexical databases such as the word lists of Obo-Manobo accounted by the SIL. The study revealed distinct variations in the Obo-Manobo language from the culture bearers of Kidapawan City, Magpet and President Roxas in the Province of North Cotabato. Obo-Manobo has lexical varieties in the way they name and describe things, person, places, events and concepts (noun and adjectives); use the main verb, prepositions and conjunctions. These are all accounted in the glossary of terms – the output of this study.
The concept of norm, developed by Eugeniu Coșeriu in 1952 as part of the trichotomy system, norm, speech, and later related to the notion of language type, was defined as a system of obligatory, common, normal actualizations and traditions of the language, which are not necessarily functional, and which vary from one community of speakers to another. In the view of the Tübingen linguist, within the same linguistic community and the same functional system more types of norms can be identified: the norm of the literary language, the norm of the vernacular, the norm of familiar language, the norm of formal language, the norm of vulgar language, etc. In what concerns norm, the Romanian linguist also makes another important distinction, namely that between social norm and individual norm. The present paper deal s with linguistic norm, as it was theorized by Eugeniu Coșeriu, and then focuses on the norm of the Romanian language in particular. Considering the “architecture” of historical language, i.e. the internal differences of the language: diatopic, diastratic and diaphasic also described and analysed by Eugeniu Coșeriu it will illustrate the division of the norm within the Romanian language.
Abstract Formal constraints on crossing dependencies have played a large role in research on the formal complexity of natural language grammars and parsing. Here we ask whether the apparent evidence for constraints on crossing dependencies in treebanks might arise because of independent constraints on trees, such as low arity and dependency length minimization. We address this question using two sets of experiments. In Experiment 1, we compare the distribution of formal properties of crossing dependencies, such as gap degree, between real trees and baseline trees matched for rate of crossing dependencies and various other properties. In Experiment 2, we model whether two dependencies cross, given certain psycholinguistic properties of the dependencies. We find surprisingly weak evidence for constraints originating from the mild context-sensitivity literature (gap degree and well-nestedness) beyond what can be explained by constraints on rate of crossing dependencies, topological properties of the trees, and dependency length. However, measures that have emerged from the parsing literature (e.g., edge degree, end-point crossings, and heads’ depth difference) differ strongly between real and random trees. Modeling results show that cognitive metrics relating to information locality and working-memory limitations affect whether two dependencies cross or not, but they do not fully explain the distribution of crossing dependencies in natural languages. Together these results suggest that crossing constraints are better characterized by processing pressures than by mildly context-sensitive constraints.
BACKGROUND: Fluorescence angiography in colorectal surgery is a technique that may lead to lower anastomotic leak rates. However, the interpretation of the fluorescent signal is not standardised and there is a paucity of data regarding interobserver agreement. The aim of this study is to assess interobserver variability in selection of the transection point during fluorescence angiography before anastomosis. METHODS: An online survey with still images of fluorescence angiography was distributed through colorectal surgery channels containing images from 13 patients where several areas for transection were displayed to be chosen by raters. Agreement was assessed overall and between pre-planned rater cohorts (experts vs non-experts; trainees vs consultants; colorectal specialists vs non colorectal specialists), using Fleiss' kappa statistic. RESULTS: 101 raters had complete image ratings. No significant difference was found between raters when choosing a point of optimal bowel transection based on fluorescence angiography still images. There was no difference between pre-planned cohorts analysed (experts vs non-experts; trainees vs consultants; colorectal specialists vs non colorectal specialists). Agreement between these cohorts was poor (<.26). CONCLUSION: Whilst there is no learning curve for the technical adoption of FA, understanding the fluorescent signal characteristics is key to successful use. We found significant variation exists in interpretation of static fluorescence angiography data. Further efforts should be employed to standardise fluorescence angiography assessment.
University level second language (L2) creative writing courses are widespread around the world. However, little empirical research has been done to document teachers and students’ experiences with courses whose main focus is to develop students’ creativity. In response, this case study was conducted to explore how an English native-speaking (NS) creative writing teacher and non-English majors at a Taiwanese public university evaluated an elective L2 creative writing course through the investigation of the teachers’ language ideology, assessment strategies, and learning activities. Data included the course syllabus, the teacher’s and the students’ reflection reports, and the students’ L2 creative writing portfolios. Through qualitative content analysis, it was found that this creative writing course did focus on teaching creative writing skills. Despite the NS identity of the teacher, in resonance with the spirit of the world Englishes pedagogical paradigm, the notion of conforming to NS linguistic norms and accuracy was not introduced to students. Lecturing, sharing, and workshopping were found to be effective learning and assessment activities, while students experienced the most difficulty with keeping a creative writing journal. We argue that creativity should and can be pursued as the main objective in English creative writing courses even by NS teachers. Lastly, implications for research and pedagogy were considered.
Liking and pleasantness are common concepts in psychological emotion theories and in everyday language related to emotions. Despite obvious similarities between the terms, several empirical and theoretical notions support the idea that pleasantness and liking are cognitively different phenomena, becoming most evident in the context of emotion regulation and art enjoyment. In this study it was investigated whether liking and pleasantness indicate behaviourally measurable differences, not only in the long timespan of emotion regulation, but already within the initial affective responses to visual and auditory stimuli. A cross-modal affective priming protocol was used to assess whether there is a behavioural difference in the response time when providing an affective rating to a liking or pleasantness task. It was hypothesized that the pleasantness task would be faster as it is known to rely on rapid feature detection. Furthermore, an affective priming effect was expected to take place across the sensory modalities and the presentative and non-presentative stimuli. A linear mixed effect analysis indicated a significant priming effect as well as an interaction effect between the auditory and visual sensory modalities and the affective rating tasks of liking and pleasantness: While liking was rated fastest across modalities, it was significantly faster in vision compared to audition. No significant modality dependent differences between the pleasantness ratings were detected. The results demonstrate that liking and pleasantness rating scales refer to separate processes already within the short time scale of one to two seconds. Furthermore, the affective priming effect indicates that an affective information transfer takes place across modalities and the types of stimuli applied. Unlike hypothesized, liking rating took place faster across the modalities. This is interpreted to support emotion theoretical notions where liking and disliking are crucial properties of emotion perception and homeostatic self-referential information, possibly overriding pleasantness-related feature analysis. Conclusively, the findings provide empirical evidence for a conceptual delineation of common affective processes.
PURPOSE: The graphic symbol is the foundation of augmentative and alternative communication (AAC) for many preliterate individuals; however, research has focused primarily on static graphic symbol sequences despite mainstream commercial technologies such as animation. The goal of this study was to compare static and animated symbol sequences across receptive communication outcome measures and psycholinguistic features (e.g., word frequency). METHOD: A counterbalanced, 2 × 2 × 2 mixed design was used to investigate the effects of symbol format (animated and static), first condition (animated or static), and first experimental task (identification or labeling) on identification accuracy and labeling accuracy of graphic symbol sequences (five symbols) in 24 children with typical development ages 7 and 8 years old. Additionally, three 2 × 2 repeated-measures analyses of variance were conducted using symbol format (animated and static) and (a) word frequency (low, high), (b) imageability (low, high), and (c) concreteness (low, high). RESULTS: In addition to superior identification and labeling accuracy of animated sequences, a significant interaction between symbol format and the first condition was observed for both experimental tasks. When the animation format was the first condition, then the children's performance improved in the subsequent static condition. Finally, word frequency, imageability, and concreteness ratings for all verbs and prepositions had significant effects on labeling accuracy of verbs and prepositions. Significant interactions between symbol format and psycholinguistic features were also found. For example, highly imageable, animated verbs were labeled with greater accuracy when compared with all other variables. CONCLUSIONS: Animation technology appears to alleviate some of the burden associated with word- and sentence-level outcomes in children with typical development. Moreover, animation appears to reduce the effects of psycholinguistic features such as word frequency and imageability by increasing the transparency of the symbol. Given the increase in research in this area, speech-language pathologists may consider adopting animated graphic symbols on a case-by-case basis as a tool to augment the learning of word classes in which movement is integral to comprehension.
Phrase-level sentiment intensity prediction is difficult due to the inclusion of linguistic modifiers (e.g., negators, degree adverbs, and modals) potentially resulting in an intensity shift or polarity reversal for the modified words. This study develops a graph-based Chinese parser based on the deep biaffine attention model to obtain dependency structures and relations. These obtained dependency features are then used in our proposed Weighted-sum Tree GRU network to predict phrase-level sentiment intensity in the valence-arousal dimensions. Dependency parsing results using the Sinica Treebank indicate that our graph-based model outperforms transition-based methods such as MLP and stack-LSTM with identical findings for English dependency parsing. Experimental results on the Chinese EmoBank indicate that our Weighted-sum Tree GRU network model outperforms other transformer-based neural networks such as BERT, ALBERT, XLNET and ELECTRA, reflecting the effectiveness of linguistic dependencies in phrase-level sentiment intensity predication tasks. In addition, our proposed model requires fewer parameters and less inference time for quantitative analysis, making the proposed model is relatively lightweight and efficient.
<span lang="EN-US">Language identification (LI) in textual documents is the process of automatically detecting the language contained in a document based on its content. The present language identification techniques presume that a document contains text in one of the fixed set of languages. However, this presumption is incorrect when dealing with multilingual document which includes content in more than one possible language. Due to the unavailability of standard corpora for Hindi-English mixed lingual language processing tasks, we propose the language lexicons, a novel kind of lexical database that augments several bilingual language processing tasks. These lexicons are built by learning classifiers over English and transliterated Hindi vocabulary. The designed lexicons possess condensed quantitative characteristics which reflect their linguistic strength in respect of Hindi and English language. On evaluating the lexicons, it is observed that words of the same language tend to cluster together and are separable over language classes. On comparing the classifier performance with existing works, the proposed lexicon models exhibit the better performance.</span>
Sentiment Analysis is very important for many applications and tasks. It became very useful tool in market trends predictions, social media monitoring, sufficient predictor for election and products. Many approaches, by many researchers, were used for this task starting from rules-based to machine learning and deep learning approaches on different types of data. All of these methods suffer from a lack of accuracy as a result of not achieving the correct understanding of the texts. In this work, an approach for extracting and using the exact meaning of the word within the context is used as part of initialization data representation. Three well-known classifiers were used (Narve Bayes, Support Vector Machine, and Maximum entropy) for testing this approach on two languages (Arabic and English). The system was evaluated using precision, recall, f-measure, and accuracy on two different datasets (Stanford Sentiment Treebank for English language and LABR-V2 for Arabic language). The results showed that the exact meaning of the words is very useful and the accuracy was increased by range 7.3% for best cases and increased by range of 4% in worst cases. The contributions of this work can be summarized by: (i) using exact meaning of the word for two languages, (ii) using new approach for adding synonyms and antonyms for exact meaning of the word in context, and (iii) better accuracy was received compared with the baseline classifiers.
Objective: To assess the impact of nasal deviation on the perception of the maxillary dental centreline position as judged by orthodontists, dentists and laypersons. Design: Cross-sectional study. Setting: Barts and the London School of Medicine and Dentistry, Queen Mary University of London, UK. Participants: Three groups of raters comprising 30 orthodontists, 30 dentists and 30 laypersons. Methods: A frontal photograph of a smiling white woman was captured and digitally manipulated with varying degrees of nasal deviation and dental centreline (DC) position in increments of 1.5 mm and 3 mm to the right and left. Three rater groups assessed the attractiveness of images using a visual analogue scale (VAS). Multiple regression analysis was undertaken, and images were compared using the Tukey HSD method. Results: Using a mixed linear model, the intraclass correlation coefficient (ICC) was estimated in the range of 69%–86%, indicating good inter-rater reliability. The interaction between image rating and nasal position ( P < 0.001), DC position ( P < 0.001) and the relationship between nose and DC position ( P < 0.001) were found to be statistically significant with symmetrical upper midline and nasal tip position, both considered to be most aesthetically pleasing. Image rating was not influenced by rater group type ( P = 0.995), age ( P = 0.983) or sex ( P = 0.476). Conclusion: There was a preference for a central and coincident nose and maxillary DC position uniformly across the rater groups. Deviations of the nose, DC and their interactions negatively impacted on perceived smile aesthetics with increasing extent and opposing direction of deviations rated progressively more unaesthetic. No differences were observed between orthodontists, general dental practitioners and lay people with respect to perceived impact on smile aesthetics.
The term “transgression” is traditionally associated with the infringement of what is prescribed. However, a closer look at its nature suggests that it is an integral part of the norm, as well as a starting point for innovation, in this case linguistic. The study focuses on the linguistic landscape (LL) of Spain, where five official languages share regional official status with Castilian Spanish. Further, these languages coexist in the LL with immigrant ones and English as an international language. In this environment, the article explores how linguistic transgression is reflected in the LL and what motivations underlie such non-normative uses. Given the spatial and grammatical limitations of the texts in the LL, the study focuses on aspects of code preference and orthography. To this end, we work on photographs taken in different Spanish regions which reflect the range of transgressive linguistic practices present in the public space. The evidence gathered allows us to suggest the grouping of these techniques under the categories of code (or variant) choice/elimination, exoticisation, re-representation and re-signification. The subsequent analysis presents linguistic transgression in LL as a voluntary, motivated and intentional social act that reflects identity, socio-cultural, but also commercial motivations. These motivations lead street-text authors to force the linguistic norm in their texts in order to claim their identity, show their solidarity with ideologies, resist linguistic policies or seek identification with their audience’s sensitivities for trade purposes.
Anxiety disorders are characterized by cognitive dysfunctions which contribute to the patient's profound disabilities. The threat of shock paradigm represents a validated psychopathological model of anxiety to measure the impact of anxiety on cognitive processes. We have developed an online version of the threat of scream paradigm (ToSP) to investigate the impact of experimental anxiety on recognition memory. Two animated passive walkthrough videos (either under threat of scream or safety conditions) were shown to healthy participants. Recognition memory, primacy vs. recency effects, and subjective estimations of the length of encoding sessions were assessed. Subjective anxiety, stress, and emotional arousal ratings indicated that experimental anxiety could successfully be induced (Safe-Threat) or reversed (Threat-Safe) between the two passive walkthrough sessions. Participants exposed to distress screams showed impaired retrieval of complex information that has been presented in an animated environment. In the threat condition, participants failed to recognize details related to the persons encountered, their spatial locations, as well as information about the temporal order and sequence of encounters. Participant groups, which received a threat announcement prior to the first walkthrough session (Threat-Threat vs. Safety-Safety and Threat-Safety vs. Safety-Threat) showed poorer recognition memory as compared to the groups that received a safety announcement (P = 0.0468 and P = 0.0426, respectively; Mann-Whitney U test, Cohen's d = 0.5071; effect size r = 0.2458). In conclusion, experimental anxiety induced by the online version of the ToSP leads to compromised recognition memory for complex multi-dimensional information. Our results indicate that cognitive functions of vulnerable populations (with limited mobility) can be evaluated online by means of the ToSP.
The main metadata projects for linguistic (language) resources developed over the past 20 years are described. These include the IMDI initiative, the OLAC metadata system. the META-SHARE meta-model, the International Standard Number of Language resources, the evaluation map of language resources, and the CLARIN component metadata model. The content of the ISO metadata standard is described. Projects for creating dictionaries, ontologies, and lexical databases for metadata of language resources are described.
Objectives/research questions: We examined stative and eventive passive bilingual compound verbs (BCVs) in Spanish/English code-switching. Of particular interest to us was the availability of passivization in bilingual eventive passive hacer “do” constructions, purportedly banned in bilingual speech due to a universal syntactic restriction. Methodology: A total of 119 bilinguals from Northern Belize and 36 from Southwest United States completed a two-alternative forced-choice acceptability task and a language background questionnaire. Data and analysis: The analysis was conducted using Thurstone’s Law of Comparative Judgment. Conclusion: For stative passive BCVs, results revealed that Spanish/English bilinguals from both contexts gave the highest ratings to code-switched constructions without the light verb hacer. For eventive passive BCVs, however, Belize bilinguals gave preferential ratings to passive constructions with the light verb hacer. Conversely, US bilinguals rejected them. Notably, among Belize bilinguals, eventive passive BCVs that were rated as most acceptable were constructions with no gender agreement between the light verb and the feminine antecedent noun. Originality: This is the first cross-community analysis that investigates stative and eventive passive BCVs in Spanish/English code-switching. Implications: Our findings show that the light verb hacer is compatible with both stative and eventive passive BCVs. Crucially, context-specific linguistic norms and social factors rather than a universal syntactic restriction primarily determine the availability of passivization in eventive passive BCVs. Our theorizing of code-switching grammars, thus, necessitates careful consideration of invariant and variable production patterns that are profoundly shaped by historical and sociolinguistic conditions.
Abstract This paper argues that Wittgenstein does not assimilate certainties to either linguistic norms or empirical propositions but assigns them to a liminal space between rule and experience. This liminal space is also brought into play in remarks written at the same time as those compiled in On Certainty, but attributed to different bodies of text ( Remarks on Colour, Last Writings on the Philosophy of Psychology ). The paper maintains that certainties express the agreement and constancy in judgements without which – as Wittgenstein contends in his Philosophical Investigations – rule-following would not be possible. It is shown that this intrinsic relation between rule-following and certainties can explain the liminal status of the latter.