Papers reviewed and determined not to be word norm studies. Use the flag icon to report errors or suggest re-inclusion.
16504 papers
Recently, neural architecture search (NAS) has emerged as a technique of growing concern in automatic machine learning (AutoML). Meanwhile, attention-based models, such as attention-based recurrent neural network, transformer-based model, etc., have been widely used in deep learning applications. However, there is no efficient NAS method that can search the architecture of attention-based model so far. To solve this problem, we propose a framework named neural architecture search for attention-based networks (NASABN) by abstracting attention-based models and extracting undefined parts of the model, including the attention layers and cells. NASABN is flexible and general enough to fit different NAS methods, which can also be transferred across different datasets. We conduct extensive experiments with NASABN using gradient descent-based methods like DARTS on Penn Treebank (PTB) and WikiText-2 (WT2) datasets respectively, and achieve competitive performance compared with the state-of-the-art methods.
In this paper we present a method for identifying and analyzing adnominal possessive constructions in 66 Universal Dependencies treebanks. We classify adpossessive constructions in terms of their morphological type (locus of marking) and present a workflow for detecting and analyzing them typologically. Based on a preliminary evaluation, the algorithm works fairly reliably in adpossessive constructions that are morphologically marked. However, it performs rather poorly in adpossessive constructions that are not marked morphologically, so-called zero-marked constructions, because of difficulties in identifying these constructions with the current annotation. We also discuss different types of variation in annotation in different treebanks for the same language and for treebanks of closely related languages. The research focuses on one well-circumscribed and universal construction in the hope of generating more interest in using UD for cross-linguistic comparison and for contributing towards developing yet more consistent annotation of constructions in the UD annotation scheme.
Summary In this work a novel framework for modeling role and task allocation in Cooperative Heterogeneous Multi‐Robot Systems (CHMRSs) is presented. This framework encodes a CHMRS as a set of multidimensional relational structures (MDRSs). This set of structure defines collaborative tasks through both temporal and spatial relations between processes of heterogeneous robots. These relations are enriched with tensors which allow for geometrical reasoning about collaborative tasks. A learning schema is also proposed in order to derive the components of each MDRS. According to this schema, the components are learnt from data reporting the situated history of the processes executed by the team of robots. Data are organized as a multirobot collaboration treebank (MRCT) in order to support learning. Moreover, a generative approach, based on a probabilistic model, is combined together with nonnegative tensor decomposition (NTD) for both building the tensors and estimating latent knowledge. Preliminary evaluation of the performance of this framework is performed in simulation with three heterogeneous robots, namely, two Unmanned Ground Vehicles (UGVs) and one Unmanned Aerial Vehicle (UAV).
Arabic dependency parsers perform poorly compared to parsers of other languages. There is little research on improving the performance of Arabic parsers. However, recent research has shown slight improvements in the performance of dependency parsers by utilizing the lexical level of a dependency treebank. To our knowledge, no previous study has studied the effect of utilizing the syntactic level. In this study, we empirically investigated the impact of varying the set of dependency relations on the performance of Arabic dependency parsers. The results were compared to those of previous studies, and showed that having an appropriate set of dependency relations could improve the performance of an Arabic dependency parser.
Sentiment analysis, especially for long documents, plausibly requires methods capturing complex linguistics structures. To accommodate this, we propose a novel framework to exploit task-related discourse for the task of sentiment analysis. More specifically, we are combining the largescale, sentiment-dependent MEGA-DT treebank with a novel neural architecture for sentiment prediction, based on a hybrid TreeLSTM hierarchical attention model. Experiments show that our framework using sentiment-related discourse augmentations for sentiment prediction enhances the overall performance for long documents, even beyond previous approaches using well-established discourse parsers trained on human annotated data. We show that a simple ensemble approach can further enhance performance by selectively using discourse, depending on the document length.
The urban sound environment is one of the layers that characterizes a city, and several methodologies are used for its assessment, including the soundwalk approach. However, this approach has been tested mainly with adults. In the work presented here, the aim is to investigate a soundwalk methodology for children, analyzing the sound environment of five different sites of Gothenburg, Sweden, from children's view-point, giving them the opportunity to take action as an active part of society. Both individual assessment of the sound environment and acoustic data were collected. The findings suggested that among significant results, children tended to rank the sound environment as slightly better when lower levels of background noise were present ( L A 90 ). Moreover, traffic dominance ratings appeared as the best predictor among the studied sound sources: when traffic dominated as a sound source, the children rated the sound environment as less good. Additionally, traffic volume appeared as a plausible predictor for sound environment quality judgments, since the higher the traffic volume, the lower the quality of the sound environment. The incorporation of children into urban sound environment research may be able to generate new results in terms of children's understanding of their sound environment. Moreover, sound environment policies can be developed from and for children.
In this paper, we present our submission for subtask A of the Common Sense Validation and Explanation (ComVE) shared task. We examine the ability of large-scale pre-trained language models to distinguish commonsense from non-commonsense statements. We also explore the utility of external resources that aim to supplement the world knowledge inherent in such language models, including commonsense knowledge graph embedding models, word concreteness ratings, and text-to-image generation models. We find that such resources provide insignificant gains to the performance of fine-tuned language models. We also provide a qualitative analysis of the limitations of the language model fine-tuned to this task.
We use Universal Dependencies treebanks to test whether a well-known typological trade-off between word order freedom and richness of morphological marking of core arguments holds within individual languages. Using Russian and German treebank data, we show that the following phenomenon (sometimes dubbed word order freezing) does occur: those sentences where core arguments cannot be distinguished by morphological means (due to case syncretism or other kinds of ambiguity) have more rigid order of subject, verb and object than those where unambiguous morphological marking is present. In ambiguous clauses, word order is more often equal to the one which is default or dominant (most frequent) in the language. While Russian and German differ with respect to how exactly they mark core arguments, the effect of morphological ambiguity is significant in both languages. It is, however, small, suggesting that languages do adapt to the evolutionary pressure on communicative efficiency and avoidance of redundancy, but that the pressure is weak in this particular respect.
Hoarding disorder has become an official disorder in the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5). Hoarding disorder affects approximately 1.5% to 5% of the general population, and there is no known literature that has examined the prevalence of hoarding disorder among homeless populations or those living in supported housing, although hoarding problems can jeopardize their housing situation. This study used the Clutter Image Rating to estimate the prevalence of possible hoarding behavior among 660 adults living in supported housing. The results indicate that 18.5% of supported housing residents had hoarding behavior, which is more than three times the prevalence reported in the general population. These results suggest that hoarding behavior and possibly hoarding disorder may be more prevalent among those with histories of homelessness and housing instability, which may be of concern because it may affect both housing and health statuses.
In this paper, we explore self-distillation as a means to improve statistical dependency parsing models for Dutch and German over purely supervised training. Self-distillation (Furlanello et al. 2018) trains a new student model on the output of an existing (weaker) teacher model. In contrast to most previous work on self-distillation, we perform distillation using a large, unannotated corpus. We show that in dependency parsing as sequence labeling (Spoustov´a and Spousta 2010, Strzyz et al. 2019), self-distillation plus finetuning provides large improvements over models that use supervised training. We carry out experiments on the German T¨uBa-D/Z universal dependency (UD) treebank (C¸ ¨oltekin et al. 2017) and the UD conversion of the Dutch Lassy Small treebank (Bouma and van Noord 2017). We find that self-distillation improves German parsing accuracy of a bidirectional LSTM parser from 92.23 to 94.33 Labeled Attachment Score (LAS). Similarly, on Dutch we see improvement from 89.89 to 91.84 LAS.
Overclaiming has been described as people’s tendency to overestimate their cognitive abilities in general and their knowledge in particular. We discuss four different perspectives on the phenomenon of overclaiming that have been proposed in the research literature: Overclaiming as a result of a) self-enhancement tendencies, b) as a cognitive bias (e.g., hindsight bias, memory bias), c) as proxy for cognitive abilities, and d) as sign of creative engagement. Moreover, we discuss two different scoring methods for an OCQ (signal detection theory vs. familiarity ratings). To distinguish between the different viewpoints of what overclaiming is, we juxtaposed overclaiming, as indicated by claiming familiarity with non-existent terms, with fluid and crystallized intelligence, self-reported knowledge, creativity, faking ability, and personality. Overclaiming was measured with a newly comprised overclaiming questionnaire. Results of several latent variable analyses based upon a multivariate study with 298 participants were: First, overclaiming is neither predicted by honesty-humility nor faking ability and therefore reflects something different than mere self-enhancement tendencies. Second, overclaiming is not predicted by crystallized intelligence, but is highly predictive of self-reported knowledge and, thus, not suitable as an index or a proxy for cognitive abilities. Finally, overclaiming is neither related to divergent thinking and originality, and only moderately predicted by self-reported openness creativity from the HEXACO which means that overclaiming does not reflect creative ability. In sum, our results favor an interpretation of overclaiming as a phenomenon that requires more than self-enhancement motivation, in contrast to the claim that was initially proposed in the literature.
In this Letter, the authors introduce a novel approach to learn representations for sentence‐level paraphrase identification (PI) using BERT and ten natural language processing tasks. Their method trains an unsupervised model called BERT with two different tasks to detect whether two sentences are in paraphrase relation or not. Unlike conventional BERT, which fine tunes the target task such as PI to pre‐trained BERT, twice fine‐tuning deep neural networks first fine tune each task (e.g. general language understanding evaluation tasks, question answering, and paraphrase adversaries from word scrambling task) and second fine tune target PI task. As a result, the multi‐fine‐tuned BERT model outperformed the fine‐tuned model only with Microsoft Research Paraphrase Corpus (MRPC), which is paraphrase data, except for one case of Stanford Sentiment Treebank ‐ 2 (SST‐2). Multi‐task fine‐tuning is a simple idea but experimentally powerful. Experiments show that fine‐tuning just PI tasks to the BERT already gives enough performance, but additionally, fine‐tuning similar tasks can affect performance (3.4% point absolute improvement) and be competitive with the state‐of‐the‐art systems.
This paper proposes a framework for the expression of typological statements which uses realvalued logics to capture the empirical truth value (truth degree) of a formula on a given data source, e.g. a collection of multilingual treebanks with comparable annotation. The formulae can be arbitrarily complex expressions of propositional logic. To illustrate the usefulness of such a framework, we present experiments on the Universal Dependencies treebanks for two use cases: (i) empirical (re-)evaluation of established formulae against the spectrum of available treebanks and (ii) evaluating new formulae (i.e. potential candidates for universals) generated by a search algorithm.
We present a parsing model for projective dependency trees which takes advantage of the existence of complementary dependency annotations for a language. This is the case for Arabic with the availability of CATiB and UD treebanks. Our system performs syntactic parsing according to both annotation types jointly as a sequence of arc-creating operations following the Easy-First approach, and partially created trees for one annotation type are also available to the other as features for the score function. This method gives error reduction of 9.9% on CATiB and 6.1% on UD compared to a single-task baseline, and ablation tests show that the main contribution of this reduction is given by sharing tree representation between tasks, and not simply sharing BiLSTM layers as is usually performed in NLP multitask systems.
This paper is concerned with semantic parsing for English as a second language (ESL). Motivated by the theoretical emphasis on the learning challenges that occur at the syntaxsemantics interface during second language acquisition, we formulate the task based on the divergence between literal and intended meanings. We combine the complementary strengths of English Resource Grammar, a linguistically-precise hand-crafted deep grammar, and TLE, an existing manually annotated ESL UD-TreeBank with a novel reranking model. Experiments demonstrate that in comparison to human annotations, our method can obtain a very promising SemBanking quality. By means of the newly created corpus, we evaluate state-of-the-art semantic parsing as well as grammatical error correction models. The evaluation profiles the performance of neural NLP techniques for handling ESL data and suggests some research directions.
Recently, due to the interplay between syntax and semantics, incorporating syntactic knowledge into neural semantic role labeling (SRL) has achieved much attention. Most of the previous syntax-aware SRL works focus on explicitly modeling homogeneous syntactic knowledge over tree outputs. In this work, we propose to encode heterogeneous syntactic knowledge for SRL from both explicit and implicit representations. First, we introduce graph convolutional networks to explicitly encode multiple heterogeneous dependency parse trees. Second, we extract the implicit syntactic representations from syntactic parser trained with heterogeneous treebanks. Finally, we inject the two types of heterogeneous syntax-aware representations into the base SRL model as extra inputs. We conduct experiments on two widely-used benchmark datasets, i.e., Chinese Proposition Bank 1.0 and English CoNLL-2005 dataset. Experimental results show that incorporating heterogeneous syntactic knowledge brings significant improvements over strong baselines. We further conduct detailed analysis to gain insights on the usefulness of heterogeneous (vs. homogeneous) syntactic knowledge and the effectiveness of our proposed approaches for modeling such knowledge.
Traditional NLP has long held (supervised) syntactic parsing necessary for\nsuccessful higher-level semantic language understanding (LU). The recent advent\nof end-to-end neural models, self-supervised via language modeling (LM), and\ntheir success on a wide range of LU tasks, however, questions this belief. In\nthis work, we empirically investigate the usefulness of supervised parsing for\nsemantic LU in the context of LM-pretrained transformer networks. Relying on\nthe established fine-tuning paradigm, we first couple a pretrained transformer\nwith a biaffine parsing head, aiming to infuse explicit syntactic knowledge\nfrom Universal Dependencies treebanks into the transformer. We then fine-tune\nthe model for LU tasks and measure the effect of the intermediate parsing\ntraining (IPT) on downstream LU task performance. Results from both monolingual\nEnglish and zero-shot language transfer experiments (with intermediate\ntarget-language parsing) show that explicit formalized syntax, injected into\ntransformers through IPT, has very limited and inconsistent effect on\ndownstream LU performance. Our results, coupled with our analysis of\ntransformers' representation spaces before and after intermediate parsing, make\na significant step towards providing answers to an essential question: how\n(un)availing is supervised parsing for high-level semantic natural language\nunderstanding in the era of large neural models?\n
Language games represent one of the most fascinating challenges of research in Artificial Intelligence. One of the problems that lies in this set is the development of an automated solution generator for Crossword puzzles. Automatic resolution of crossword puzzles is an open natural language challenge that demands the filling of puzzle grid with candidate answers, while satisfying the grid constraints. Moreover, providing an accurate list of answer candidates has a major impact on the performance of the automatic crossword resolution task. This paper proposes a semantic-based search approach for generating candidate answer lists for definition type clues by taking advantage of the lexical relations encoded in WordNet, the lexical database for English.
The article examines the problem of the nomination of regional vocabulary, to a certain extent reflecting the peculiarities of the life of the inhabitants of a particular region. The relevance of the work is due to the heightened interest in the study of territorially marked linguistic units operating in cities in recent decades. Their study is extremely important not only for linguistic regional studies, but also for determining the peculiarities of the development of a literary language that actively interacts with regions. In this regard, the problem of designating the linguistic units of the regiolect arises, which has not yet been fully solved. The heterogeneous nature of the regiolects (dialects, geographically marked vernaculars, literary linguistic units, local borrowings), contradictory interpretation of the linguistic norm, ignorance of the systemic nature of terminology complicate the aforementioned problem. The purpose of the article is to determine the most appropriate term to denote territorially marked vocabulary. To achieve this goal, the following research methods were used: - theoretical - analysis and synthesis of linguistic literature on the research problem; - comparative-comparative - comparison of terms used to denote regional vocabulary, identifying the degree of their consistency and adequacy to the nominated object. In the course of the work done, the following results were obtained. An extraordinary inconsistency in the terminological naming of territorially marked vocabulary is revealed. The advantages and disadvantages of each term are established, taking into account the accuracy of the designated object and the systemic nature of the terminology. The term is defined that most adequately reflects the essence of the designated object and is organically included in the system of linguistic terminology. The relation of regionalisms to general literary vocabulary is revealed and their non-standardization is proved. The inconsistency of identifying the regional linguistic norm and the regional variant of the literary language has been substantiated.
In this paper, a novel structured pruning approach for learning efficient long short-term memory (LSTM) network architectures is proposed. More specifically, the eigenvalues of the covariance matrix associated with the responses of each LSTM layer are computed and utilized to quantify the layers' redundancy and automatically obtain an individual pruning rate for each layer. Subsequently, a Geometric Median based (GM-based) criterion is used to identify and prune in a structured way the most redundant LSTM units, realizing the pruning rates derived in the previous step. The experimental evaluation on the Penn Treebank text corpus and the large-scale YouTube-8M audio-video dataset for the tasks of word-level prediction and visual concept detection, respectively, shows the efficacy of the proposed approach <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup>.
Advertising can be regarded as a means of persuading consumers to take action with respect to products, ideas or services. It is a complex phenomenon because it involves two interacting processes, namely, communication and persuasion, and both processes are many-faceted. When advertisers intend to make the advertisement language creative and inventive, they use language different from the linguistic norm. Being unconventional, the advertisements which adopt deviation can give the readers unexpected surprise and make a strong impression on their mind. This kind of language phenomenon is technically called linguistic deviation. Then based on some scholars’ classifications, the author classifies different deviations in advertising (deviation of phonology, graphology, lexicon, and grammar) and illustrates each deviation with some typical examples.
Using data from the World Atlas of Language Structures and the Universal Dependencies treebanks, we provide converging evidence from linguistic typology and comparative corpus linguistics for an efficiency-based trade-off in the encoding of referentially accessible subjects. Specifically, when familiar subjects are marked as bound elements attaching to the verb, the chances of having obligatory independent subject pronouns decrease significantly across the world's languages. At the same time, there is a trend against not encoding the subject at all, leading us to postulate an overall tendency to encode familiar subjects once and only once in a neutral topiccomment utterance. This tendency is mirrored in more fine-grained corpus data from Slavic: East Slavic languages, in contrast to the other members of the genus, have past forms without verbal subject encoding, and it is precisely with these (former participle) forms that the use of independent subject pronouns is significantly higher than with other, non-participial verb forms. By contrast, the occurrence of independent subject pronouns does not differ across various verb forms in other Slavic languages, as none of them has been affected by a loss of verbal subject encoding.
The television ratings provide an effective way to analyze the popularity of TV programs and audiences’ watching habits. Most previous studies have analyzed the ratings from a single perspective. Few efforts have integrated analysis from different perspectives and explored the reasons for changes in ratings. In this paper, we design a visual analysis system called TVseer to analyze audience ratings from three perspectives: TV channels, TV programs, and audiences. The system can help users explore the factors that affect ratings, and assist them in decisions about program productions and schedules. There are six linked views in TVseer: the channel ratings view and program ratings view show ratings change information from the perspective of TV channels and programs respectively; the overlapping program competition view and the same-type program competition view indicate the competitive relationships among programs; the audience transfer view shows how audiences are moving among different channels; the audience group view displays audience groups based on their watching behavior. Besides, we also construct case studies and expert interviews to prove our system is useful and effective.
本研究依据以谓词为核心的块依存语法构建块依存树库,在句内和句间寻找谓词所支配的组块,利用汉语中组块和组块间的依存关系补全缺省部分,明确谓词支配关系。目前共标注2199篇文本,涵盖百科、新闻两个领域,共约187万字语料。本文简述了块依存语法的原则,并对组块及其依存关系进行了定义。将详细介绍标注流程、标注一致率、数据分布等情况。基于现有的树库,本研究发现汉语中有约25%的小句是非自足的,约有88%的核心谓词可支配1~3个从属成分。
Contextual information can modulate the conditioned response to a threat signal (conditioned stimulus, CS+): fear responses are either potentiated or attenuated depending on whether the context is threatening or safe. In this study, we investigated the influence of context on conditioned fear as well as on generalization of conditioned fear. Thirty-two participants underwent a cue-in-context learning protocol in virtual reality (VR). On Day 1 (acquisition), participants received a mild painful electric shock (unconditioned stimulus, US) in one virtual room (fear context, CTX+) at the offset of one colored light (CS+), but never at the offset of another colored light (CS-). In a second room (safety context, CTX-), the two lights were also presented, but not the US. Successful cue conditioning was indicated by aversive ratings and startle potentiation but not skin conductance responses (SCR) to CS+ versus CS- in CTX+ and not in CTX-. On Day 2 (generalization), participants re-visited both fear and safety contexts plus a generalization context (G-CTX), which was an equal mix of CTX+ and CTX-. The two CSs were shown again in all three contexts. Generalization of conditioned fear was revealed in affective ratings (CS+ was rated more aversive than CS- in G-CTX), but not in physiological measures (equal startle potentiation to CS+ versus CS- in all contexts). In sum, contextual information modulates the responses to a threat signal such that a safety context can inhibit conditioned fear. Interestingly, generalization processes also depend on contextual information.
The large communication cost for exchanging gradients between different nodes significantly limits the scalability of distributed training for large-scale learning models. Motivated by this observation, there has been significant recent interest in techniques that reduce the communication cost of distributed Stochastic Gradient Descent (SGD), with gradient sparsification techniques such as top-k and random-k shown to be particularly effective. The same observation has also motivated a separate line of work in distributed statistical estimation theory focusing on the impact of communication constraints on the estimation efficiency of different statistical models. The primary goal of this paper is to connect these two research lines and demonstrate how statistical estimation models and their analysis can lead to new insights in the design of communication-efficient training techniques. We propose a simple statistical estimation model for the stochastic gradients which captures the sparsity and skewness of their distribution. The statistically optimal communication scheme arising from the analysis of this model leads to a new sparsification technique for SGD, which concatenates random-k and top-k, considered separately in the prior literature. We show through extensive experiments on both image and language domains with CIFAR-10, ImageNet, and Penn Treebank datasets that the concatenated application of these two sparsification methods consistently and significantly outperforms either method applied alone.
Eating is a process that involves all senses. Recent research has shown that both food-intrinsic and extrinsic sensory factors play a role in the taste of the food we consume. Moreover, many studies have explored the relationship between emotional state and taste perception, where certain emotional states have been shown to alter the perception of basic tastes. This opens up a whole new world of possibilities for the design of eating environments which take into account both sensory attributes as well as their emotional associations. Here, we used virtual reality to study the effect of colours and music, with specific emotional associations, on the evaluation of cold brew coffee. Based on an online study (N=76), two colours and two pieces of music with similar emotional arousal but opposing valence ratings were chosen to produce a total of eight virtual coloured environments. Forty participants were recruited for the on-site experiment, which consisted of three blocks. First, a blind tasting of four coffee samples (0%, 2.5%, 5%, 7.5% sucrose) was carried out. Next, participants experienced the eight environments via an HTC Vive Pro headset and evaluated their expected liking, sweetness and bitterness of a mug of coffee presented in VR. Finally, they tasted identical 5% coffee samples in the same eight environments. Results revealed One of the key findings of this study that, when only one factor (colour or music) was manipulated, background colour significantly influenced coffee liking. When colour and music were used in combination, however, we found an overall effect of music valence on coffee sweetness, as well as an interaction effect of colour and music on liking. These results reinforce the importance of the extrinsic sensory and emotion factors on food expectations and liking. Overall, these results are in line with previous research, where positive emotions can lead to increased food liking and higher sweetness compared to negative emotions.
In this paper, we compare four state-of-the-art neural network dependency parsers for the Semitic language Amharic. As Amharic is a morphologically-rich and less-resourced language, the out-of-vocabulary (OOV) problem will be higher when we develop data-driven models. This fact limits researchers to develop neural network parsers because the neural network requires large quantities of data to train a model. We empirically evaluate neural network parsers when a small Amharic treebank is used for training. Based on our experiment, we obtain an 83.79 LAS score using the UDPipe system. Better accuracy is achieved when the neural parsing system uses external resources like word embedding. Using such resources, the LAS score for UDPipe improves to 85.26. Our experiment shows that the neural networks can learn dependency relations better from limited data while segmentation and POS tagging require much data.
In this paper we present work carried out for the Ac-ComplIt task. ItVENSES is a system for syntactic and semantic processing that is based on the parser for Italian called ItGetaruns to analyse each sentence. In previous EVALITA tasks we only used semantics to produce the results. In this year EVALITA, we used both a statistically based approach and the semantic one used previously. The statistic approach is characterized by the use of trigrams of constituents computed by the system and checked against a trigram model derived from the constituency version of VIT – Venice Italian Treebank. Results measured in term of a correlation, are not particularly high, below 50% the Acceptability task and slightly over 30% the Complexity one.
Automatic speech recognition and spoken dialogue systems have made great advances through the use of deep machine learning methods. This is partly due to greater computing power but also through the large amount of data available in common languages, such as English. Conversely, research in minority languages, including sign languages, is hampered by the severe lack of data. This has led to work on transfer learning methods, whereby a model developed for one language is reused as the starting point for a model on a second language, which is less resourced. In this paper, we examine two transfer learning techniques of fine-tuning and layer substitution for language modelling of British Sign Language. Our results show improvement in perplexity when using transfer learning with standard stacked LSTM models, trained initially using a large corpus for standard English from the Penn Treebank corpus
BACKGROUND: Traumatic experiences are associated with neurofunctional dysregulations in key regions of the emotion regulation circuits. In particular, amygdala responsivity to negative stimuli is exaggerated while engagement of prefrontal regulatory control regions is attenuated. Successful application of emotion regulation (ER) strategies may counteract this disbalance, however, application of learned strategies in daily life is hampered in individuals afflicted by posttraumatic stress disorder (PTSD). We hypothesized that a single session of real-time fMRI (rtfMRI) guided upregulation of prefrontal regions during an emotion regulation task enhances self-control during exposure to negative stimuli and facilitates transfer of the learned ER skills to daily life. METHODS: In a cross-over design, individuals with a PTSD diagnosis after a single traumatic event (n = 20) according to DSM-IV-TR criteria and individuals without a formal psychiatric diagnosis (n = 21) underwent a cognitive reappraisal training. In randomized order, all participants completed two rtfMRI neurofeedback (NF) runs targeting the left lateral prefrontal cortex (lPFC) and two control runs without NF (NoNF) while using cognitive reappraisal to reduce their emotional response to negative scenes. During the NoNF runs, two %%-signs were displayed instead of the two-digit feedback (FB) to achieve a comparable visual stimulation. The project aimed at defining the clinical potential of the training according to three success markers: (1) NF induced changes in left lateral prefrontal cortex and bilateral amygdala activity during the regulation of aversive scenes compared to cognitive reappraisal alone (primary registered outcome), (2) associated changes on the symptomatic and behavioral level such as indicated by PTSD symptom severity and affect ratings, (3) clinical utility such as indicated by perceived efficacy, acceptance, and transfer to daily life measured four weeks after the training. RESULTS: In comparison to the reappraisal without feedback, a neurofeedback-specific decrease in the left lateral PFC (d = 0.54) alongside an attenuation of amygdala responses (d = 0.33) emerged. Reduced amygdala responses during NF were associated with symptom improvement (r = -0.42) and less negative affect (r = -0.63) at follow-up. The difference in symptom scores exceeds requirements for a minimal clinically important difference and corresponds to a medium effect size (d = 0.64). Importantly, 75% of individuals with PTSD used the strategies in daily life during a one-month follow-up period and perceived the training as efficient. CONCLUSION: Our findings suggest beneficial effects of the NF training indicated by reduced amygdala responses that were associated with improved symptom severity and affective state four weeks after the NF training as well as patient-centered perceived control during the training, helpfulness and application of strategies in daily life. However, reduced prefrontal involvement was unexpected. The study suggests good tolerability of the training protocol and potential for clinical use in the treatment of PTSD.
OBJECTIVES: The objective of this paper was to examine the implementation and effectiveness of a community-based intervention for hoarding disorder (HD) using Cognitive Rehabilitation and Exposure/Sorting Therapy (CREST). DESIGN: This was a mixed-method, pre-post quasi-experimental study informed by the Practical, Robust Implementation and Sustainability Model for implementation science. SETTING: Program activities took place in San Diego County, mainly within clients' homes or community, with some activities in-office. PARTICIPANTS: Participants were aged 60 years or older, met eligibility for Medi-Cal or were uninsured, and met criteria for HD. INTERVENTION: A manualized, mobile protocol that incorporated CREST was utilized. MEASUREMENTS: The Clutter Image Rating and Hoarding Rating Scale were used as effectiveness outcomes. An investigator-created staff questionnaire was used to evaluate implementation. RESULTS: Thirty-seven clients were reached and enrolled in treatment and 15 completed treatment during the initial 2 years of the program. There were significant changes in hoarding severity and clutter volume. Based on the initial 2 years of the program, funding was provided for expansion to cover additional San Diego County regions and hire more staff clinicians in year three. CONCLUSION: Preliminary data suggest that the CREST intervention can be successfully implemented in a community setting with positive results for older adults with HD.
Deep neural networks (DNNs) have become the gold standard for solving challenging classification problems, especially given complex sensor inputs (e.g., images and video). While DNNs are powerful, they are also brittle, and their inner workings are not fully understood by humans, leading to their use as "black-box" models. DNNs often generalize poorly when provided new data sampled from slightly shifted distributions; DNNs are easily manipulated by adversarial examples; and the decision-making process of DNNs can be difficult for humans to interpret. To address these challenges, we propose integrating DNNs with external sources of semantic knowledge. Large quantities of meaningful, formalized knowledge are available in knowledge graphs and other databases, many of which are publicly obtainable. But at present, these sources are inaccessible to deep neural methods, which can only exploit patterns in the signals they are given to classify. In this work, we conduct experiments on the ADE20K dataset, using scene classification as an example task where combining DNNs with external knowledge graphs can result in more robust and explainable models. We align the atomic concepts present in ADE20K (i.e., objects) to WordNet, a hierarchically-organized lexical database. Using this knowledge graph, we expand the concept categories which can be identified in ADE20K and relate these concepts in a hierarchical manner. The neural architecture we present performs scene classification using these concepts, illuminating a path toward DNNs which can efficiently exploit high-level knowledge in place of excessive quantities of direct sensory input. We hypothesize and experimentally validate that incorporating background knowledge via an external knowledge graph into a deep learning-based model should improve the explainability and robustness of the model.
In this study we evaluate the convergent validity of a new graphical self-report tool (the EmojiGrid) for the affective appraisal of perceived touch events. The EmojiGrid is a square grid labeled with facial icons (emoji) showing different levels of valence and arousal. The EmojiGrid is language independent and efficient (a single click suffices to report both valence and arousal), making it a practical instrument for studies on affective appraisal. We previously showed that participants can intuitively and reliably report their affective appraisal (valence and arousal) of visual, auditory and olfactory stimuli using the EmojiGrid, even without additional (verbal) instructions. However, because touch events can be bidirectional and dynamic, these previous results cannot be generalized to the touch domain. In this study, participants reported their affective appraisal of video clips showing different interpersonal (social) and object-based touch events, using either the validated 9-point SAM (Self-Assessment Mannikin) scale or the EmojiGrid. The valence ratings obtained with the EmojiGrid and the SAM are in excellent agreement. The arousal ratings show good agreement for object-based touch and moderate agreement for social touch. For social touch and at more extreme levels of valence, the EmojiGrid appears more sensitive to arousal than the SAM. We conclude that the EmojiGrid can also serve as a valid and efficient graphical self-report instrument to measure human affective response to a wide range of (possibly mediated) tactile signals.
Alexithymia is a personality trait characterized by difficulties identifying and describing feelings (DIF and DDF) and an externally-oriented thinking style (EOT). The primary aim of the present study was to investigate links between alexithymia and the evaluation of emotional scenes. We also investigated whether viewers’ evaluations of emotional scenes were better predicted by specific alexithymic traits or by individual differences in sensory processing sensitivity (SPS). Participants (N = 106) completed measures of alexithymia and SPS along with a task requiring speeded judgments of the pleasantness of 120 moderately arousing scenes. We did not replicate laterality effects previously described with the scene perception task. Compared to those with weak alexithymic traits, individuals with moderate-to-strong alexithymic traits were less likely to classify positively valenced scenes as pleasant, and less likely to classify scenes with (vs. without) implied motion in a way that was consistent with normative scene valence ratings. In addition, regression analyses confirmed that reporting strong EOT and a tendency to be easily overwhelmed by busy sensory environments negatively predicted classification accuracy for positive scenes, and that both DDF and EOT negatively predicted classification accuracy for scenes depicting implied motion. These findings highlight the importance of accounting for stimulus characteristics and individual differences in specific traits associated with alexithymia and SPS when investigating the processing of emotional stimuli. Learning more about the links between these individual difference variables may have significant clinical implications, given that alexithymia is an important, transdiagnostic risk factor for a wide range of psychopathologies.
abstract Reaction times for a translation recognition study are reported where novice to expert English–ASL bilinguals rejected English translation distractors for ASL signs that were related to the correct translations through phonology, semantics, or both form and meaning (diagrammatic iconicity). Imageability ratings of concepts impacted performance in all conditions; when imageability was high, participants showed interference for phonologically related distractors, and when imageability was low participants showed interference for semantically related distractors, regardless of proficiency. For diagrammatically related distractors high imageability caused interference in experts, but low imageability caused interference in novices. These patterns suggest that imageability and diagrammaticity interact with proficiency – experts process diagrammatic related distractors phonologically, but novices process them semantically. This implies that motivated signs are dependent on the entrenchment of language systematicity; rather than decreasing their impact on language processing as proficiency grows, they build on the original benefit conferred by iconic mappings.
We tackle implicit discourse relation classification, a task of automatically determining semantic relationships between arguments. The attention-worthy words in arguments are crucial clues for classifying the discourse relations. Attention mechanisms have been proven effective in highlighting the attention-worthy words during encoding. However, our survey shows that some inessential words are unintentionally misjudged as the attention-worthy words and, therefore, assigned heavier attention weights than should be. We propose a penalty-based loss re-estimation method to regulate the attention learning process, integrating penalty coefficients into the computation of loss by means of overstability of attention weight distributions. We conduct experiments on the Penn Discourse TreeBank (PDTB) corpus. The test results show that our loss re-estimation method leads to substantial improvements for a variety of attention mechanisms, and it obtains highly competitive performance compared to the state-of-the-art methods.
Gender can be considered an embodied social concept encompassing biological as well as cultural components. In this paper, 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 sociocultural features (e.g., discrimination, politics, power), whereas Dutch participants focused more on the corporeal sphere (e.g., hormones, breasts, genitals). Study 2 replicated this finding focusing on Italian and Dutch and using a typicality rating task: sociocultural 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 Study 1 and 2 showing that Dutch participants endorsed more essentialist beliefs about gender compared to Italian participants. Our results provide evidence that gender is conceptualized differently by diverse groups more in line with sociocultural constructivist accounts and is adapted to specific cultural and linguistic environments.