Papers reviewed and determined not to be word norm studies. Use the flag icon to report errors or suggest re-inclusion.
16504 papers
Head-driven phrase structure grammar (HPSG) enjoys a uniform formalism representing rich contextual syntactic and even semantic meanings. This paper makes the first attempt to formulate a simplified HPSG by integrating constituent and dependency formal representations into head-driven phrase structure. Then two parsing algorithms are respectively proposed for two converted tree representations, division span and joint span. As HPSG encodes both constituent and dependency structure information, the proposed HPSG parsers may be regarded as a sort of joint decoder for both types of structures and thus are evaluated in terms of extracted or converted constituent and dependency parsing trees. Our parser achieves new state-of-the-art performance for both parsing tasks on Penn Treebank (PTB) and Chinese Penn Treebank, verifying the effectiveness of joint learning constituent and dependency structures. In details, we report 96.33 F1 of constituent parsing and 97.20\% UAS of dependency parsing on PTB.
Building a treebank from scratch can easily be an elaborate, highly time consuming task, especially when working with a minority language with moderately complex morphology and no existing resources. It is also then typically true that language experts and informants with suitable skill sets are a very scarce resource. In this experiment I have attempted to work in parallel on building NLP resources while gathering and annotating the treebank. In particular, I aim to build a decent coverage morphologically annotated lexicon suitable for rule-based morphological analysis as well as accompanying rules for basic morphosyntactic analysis. I propose here a workflow, that I have found useful in avoiding redoing same work with related NLP resource construction.
We introduce the first German treebank for Twitter microtext, annotated within the framework of Universal Dependencies. The new treebank includes over 12,000 tokens from over 500 tweets, independently annotated by two human coders. In the paper, we describe the data selection and annotation process and present baseline parsing results for the new testsuite.
The article aims to be an introduction to the dependency treebanks currently available for Ancient Greek and Latin, i.e., the Ancient Greek and Latin Dependency Treebank (AGLDT), the Index Thomisticus Treebank (IT-TB), the PROIEL Treebank, and the SEMATIA Treebank.Their pipelines for creation of morphosyntactic annotations are presented so as to highlight major commonalities and differences.All treebanks share the same basic underlying formalism, whereby syntactic words are connected to each other to form labeled directed acyclic graphs, and their annotation schemes, although different, are comparable to a very large extent. An introduction to the dependency treebank formalismA dependency treebank is a corpus containing a symbolic representation of the syntax of one or more texts.It can be defined as a set of sentences parsed according to the linguistic formalism of dependency grammar.Most treebanks for Ancient Greek and Latin, i.e., the Ancient Greek and Latin Dependency Treebank (AGLDT), the Index Thomisticus Treebank (IT-TB), the PROIEL Treebank, and the SEMATIA Treebank, are dependency treebanks.Even if, in the present article, I deal only with dependency treebanks, most of what follows in the present section could also be applied, mutatis mutandis, to describe constituency treebanks, such as the Nestle 1904 and SBNLGT Treebanks, 1 the major difference being that in dependency treebanks all nodes except the ROOT node are paired with tokens, 2 while in constituency treebanks nonterminal nodes, which represent phrases, such as VPs or PPs, are also licensed.The parsed sentences in a treebank are formally represented as labeled directed acyclic graphs, where each token, excluding the ROOT node, is annotated
We report on the conversion of the Hamburg Dependency Treebank The HDT consists of more than 200.000 sentences annotated with dependency structure, making every attempt at manual conversion or manual post-processing extremely costly. The conversion employs an unranked tree transducer. This formalism allows to express transformation rules in a concise way, guarantees the well-formedness of the output and is predictable to the rule writers. Together with the release of a converted subset of the HDT spanning 3 million tokens, we release an interactive workbench for writing and refining tree transducer rules. Our conversion achieves a very high labeled accuracy with respect to a manually converted gold standard of 97.3%. Up to now, the conversion effort took about 1000 hours of work.
In this paper we describe the early stage application of the Universal Dependencies to an Italian corpus from social media developed for shared tasks related to irony and stance detection.The development of this novel resource (TWITTIR Ò-UD) serves a twofold goal: it enriches the scenario of treebanks for social media and for Italian, and it paves the way for a more reliable extraction of a larger variety of morphological and syntactic features to be used by sentiment analysis tools.On the one hand, social media texts are especially hard to parse and the limited amount of resources for training and testing NLP tools further damages the situation.On the other hand, we thought that adding the Universal Dependencies format to the fine-grained annotation for irony, that was previously applied on TWITTIR Ò, might meaningfully help in the investigation of possible relationships between syntax and semantics of the uses of figurative language, irony in particular.
Head-driven phrase structure grammar (HPSG) enjoys a uniform formalism representing rich contextual syntactic and even semantic meanings. This paper makes the first attempt to formulate a simplified HPSG by integrating constituent and dependency formal representations into head-driven phrase structure. Then two parsing algorithms are respectively proposed for two converted tree representations, division span and joint span. As HPSG encodes both constituent and dependency structure information, the proposed HPSG parsers may be regarded as a sort of joint decoder for both types of structures and thus are evaluated in terms of extracted or converted constituent and dependency parsing trees. Our parser achieves new state-of-the-art performance for both parsing tasks on Penn Treebank (PTB) and Chinese Penn Treebank, verifying the effectiveness of joint learning constituent and dependency structures. In details, we report 95.84 F1 of constituent parsing and 97.00% UAS of dependency parsing on PTB.
Abstract This chapter discusses the positioning of Belarus in the international context of socioeconomic development based on an assessment of the country's dynamics in world rankings. The country's presence in the recognized world rankings and its holding high positions in them is an obvious advantage for achieving a favorable investment image. Ratings characterize the country's comparative position at the international level in a number of areas: from credit capacity to human capital development. There has been analyzed the position of the Republic of Belarus in several recognized international comparisons, such as Human Development Index, Doing Business, ICT Development Index, Global Innovation Index, Sustainable Development Goals Index, Corruption Perceptions Index, Rule of Law Index, Worldwide Governance Indicators, and others. However, Belarus is not yet participating in the international competitiveness assessment through such popular international ratings as Global Competitiveness Index and Global Entrepreneurship Monitor. The research findings show that the strongest aspects of the socioeconomic development of Belarus are in place due to the high educational level of the human capital development, gender equality, and the implementation of the UN sustainable development goals. The analysis also shows that the weaknesses of institutional environment and public administration do not enable the full implementation of the planned goals of socioeconomic development.
Machine-learning models have demonstrated great success in learning complex patterns that enable them to make predictions about unobserved data. In addition to using models for prediction, the ability to interpret what a model has learned is receiving an increasing amount of attention. These interpretations have found a number of uses, ranging from providing scientific insight to auditing the predictions themselves to ensure fairness with respect to protected categories like race or gender. However, there is still considerable confusion about the notion of interpretability. In particular, it is currently unclear both what it means to interpret a ML model, and how to actually do so.In the first part of this thesis, we address the foundational question of what it means to interpret a ML model. In particular, it is currently unclear what it means to be interpretable, and how to select, evaluate, or even discuss, methods for producing interpretations of machine-learning models. We aim to clarify these concerns by defining interpretable machine learning and constructing a unifying framework for existing methods which highlights the under-appreciated role played by human audiences. Within this framework, methods are organized into two classes: model-based and post-hoc. To provide guidance in selecting and evaluating interpretation methods, we introduce three desiderata: predictive accuracy, descriptive accuracy, and relevancy. Using our framework, we review existing work, grounded in real-world studies which exemplify our desiderata, and suggest directions for future work.The second through fourth parts introduce a succession of methods for interpreting predictions made by neural networks. The second part focuses on Long Short Term Memory networks (LSTMs) trained on question-answering and sentiment analysis, two popular tasks in natural language processing. By decomposing the LSTM's update equations, we introduce a novel method for computing feature importance scores of specific inputs for determining the output of an LSTM. In order to verify the output of our method, we use the introduced scores to search for consistently important patterns of words learned by state ofthe art LSTMs on sentiment analysis and question answering. This representation is then quantitatively validated by using the extracted phrases to construct a simple, rule-based classifier which approximates the output of the LSTM.While feature importance scores are helpful in understanding a model's predictions, they ignore the complex interactions between variables typically learned by neural networks. To this end, the third part introduces contextual decomposition (CD), an interpretation algorithm for analysing individual predictions made by standard LSTMs, without any changes to the underlying model. By decomposing the output of a LSTM, CD captures the contributions of combinations of words or variables to the final prediction of an LSTM. On the task of sentiment analysis with the Yelp and Stanford Sentiment Treebank (SST) data sets, we show that CD is able to reliably identify words and phrases of contrasting sentiment, and how they are combined to yield the LSTM's final prediction. Using the phrase-level labels in SST, we also demonstrate that CD is able to successfully extract positive and negative negations from an LSTM, something which has not previously been done.When considering interactions between variables, the number of interactions quickly becomes too large for manual inspection, leading to the question of how to automatically select and display an informative subset. In the fourth part, we introduce the use of hierarchical interpretations to explain DNN predictions through our proposed method: agglomerative contextual decomposition (ACD). Given a prediction from a trained DNN, ACD produces a hierarchical clustering of the input features, along with the contribution of each cluster to the final prediction. This hierarchy is optimized to identify clusters of features that the DNN learned are predictive. We introduce ACD using examples from Stanford Sentiment Treebank and ImageNet, in order to diagnose incorrect predictions, identify dataset bias, and extract polarizing phrases of varying lengths. Through human experiments, we demonstrate that ACD enables users both to identify the more accurate of two DNNs and to better trust a DNN's outputs. We also find that ACD's hierarchy is largely robust to adversarial perturbations, implying that it captures fundamental aspects of the input and ignores spurious noise.
Research in adults suggests that certain sleep characteristics prior to trauma exposure predicts maladaptive post-trauma outcomes. However, research has yet toconsider the influence of premorbid sleep in influencing child responses to trauma. This study included 66 pre-pubertal children residing in Houston, TX several years before and during Hurricane Harvey. At time 1 (ages 7-11) children completed one week of actigraphy and one night of polysomnography. Approximately 3 years later, 6-9 months after the hurricane, youth reported on their hurricane exposure/stress (e.g., displacement, feelings of safety) and provided reports of their emotional reactions (valence and arousal) to a series of hurricane-related photos during an in-lab experimental task. Multiple regressions controlling for the time between assessments indicated less non-rapid eye movement (NREM) stage 3 sleep was associated with greater stress during the hurricane (p =.01) and marginally greater negative valence in response to flood images (p =.098). Longer childhood sleep duration (p =.01) and sleep onset latency (SOL; p =.02) were both associated with greater exposure to hurricane-related stressors. An interaction with REM sleep (p =.005) suggested hurricane-related stress was associated with greater reactivity to images (higher negative valence ratings) but only for those with greater REM sleep (p =.004). A similar interaction emerged where greater exposure was associated with greater reactivity for those with high REM (p =.03). Similarly, exposure to hurricane-related stressors was associated with higher arousal, but only for those with higher pre-hurricane SOL (p =.03). Johnson-Neyman results suggest that this effect was significant for those with a SOL of approximately 40 minutes or longer. Consistent with findings in adults, specific sleep characteristics in childhood appear to increase risk for maladaptive responses to trauma during the sensitive, transitional period of adolescence. These findings support the role of sleep as a possible prevention target for those at-risk for experiencing trauma. National Science Foundation 1813574 (PI: Palmer); National Institutes of Health K23MH081188 (PI: Alfano); and National Institutes of Health R21MH099351 (PI: Alfano)
The aim was to investigate whether a computer-based evaluative conditioning (EC) procedure using positive social feedback is effective in enhancing body satisfaction. Prior findings in three small-scale studies were mixed showing positive effects in pre-clinical samples, but not in a clinical sample of eating disorder patients. Therefore, our main goal was to replicate the original finding of Martijn et al. (2010) in a well-powered unselected sample of 129 female undergraduates. We assessed the impact of EC on questionnaire measures of body satisfaction as well as on affective ratings of the participants’ body pictures used in the task to verify whether the EC procedure was effective in heightening the subjective valence of these pictures. Supporting the validity of the current EC procedure, participants in the experimental condition rated their own pictures after the training as more positive than participants in the control condition. However, this effect of the EC procedure did not transfer to the self-report indices of body satisfaction. In addition, women with relatively high body concern did not profit more from the EC procedure than women with relatively low body concern. Together, these findings suggest that the EC procedure in its present form is not ready for use as an intervention for improving body satisfaction. However, it seems worthwhile to investigate in future studies how the EC procedure can be further strengthened in a way that effects on stimulus ratings eventually also ‘spill over’ to the level of self-reported body satisfaction.
Sentiment Analysis is an application of Natural Langue Processing to analyze social media corpora to extract insights of corpora. Sentiment analytical results are the real feedback of the customers, which enables the organizations and companies to take appropriate decision on their products and business policies. Stemming plays in-evitable and vital role in sentiment analysis. Stemming is one of the phase of preprocessing the social media corpora. Today most of the researches uses strong stemmers to identify stem words of social media corpora. The most popular stemming algorithms such as Lancaster and Porter stemming algorithms causes prejudiced the meaning of the words. The over-stemmed words mislead the sentiment classification process. To prevent the over-stemming the Unprejudiced lighter stemming algorithm is proposed to sustain the meaning of the stemmed words. The propose Un-prejudiced algorithm uses lexical database and Parts of speech of Python Natural Language Tool Kit. There are a few stemming algorithm accuracy evaluation methods, in this paper we focused on Paice Error-rate relative to truncation (ERRT) measure to evaluate the accuracy of Lancaster, Porter and Unprejudiced stemming algorithms. The experiments were conducted on 25,758 source words and results were evaluated using Paice stem evaluation method and Sirsat method. The Paice Evaluation ERRT values 0.47209, 0.28703, 0.15502 of Lancaster, Porter, Unprejudiced respectively are proved that the Unprejudiced stemmer is more accurate than Lancaster and Porter. Sirsat’s stem evaluation method Average Words Conflation Factor (AWCF) results 10310.31, 14031.17, 23349.87 of Lancaster, Porter, Unprejudiced respectively are also proved the Unprejudiced stemming algorithm is more accurate than Lancaster and Porter stemming algorithms.
Many government schemes were unsuccessful because lack of proper feedback on the ongoing schemes, where billion dollars investment is going to be in vain. Sentiment analysis is one of best approach to analyse opinions of the peoples on various government schemes. Sentiment analysis and machine learning techniques emerged to analyse huge social media corpora to track people's views on government policies, products and services. Sentiment analysis process consists of various phases which include data discovery, data collection, data pre-processing, and data analysis. Stemming is a process to generate the morphemes in natural language sentences for various applications such as sentiment analysis, information retrieval, and domain analysis. The stemming process involved two major errors, which are over-stemming and under-stemming errors. Most of sentiment analysis natural languages processing applications used Lancaster and Porter stemming algorithms where more than one word inflected into same morpheme, which causes the etymology behaviour of the stemming word and prone to classify the tweets false positives and false negative. The proposed un-prejudice light stemming algorithm prevent etymology behaviour of morpheme and sustain its meaning during stemming process by selecting a word which has maximum number of synonyms in lexical database.
The similarity between words constitutes significant support to tasks in natural language processing. Several works use Lexical resources such as WordNet for semantic similarity and synonym identification. Nevertheless, words out-of-vocabulary or missing links between senses are perceived problems of this approach. Distributional-based proposals like word embeddings have successfully been used to meet such problems, but the lack of contextual information can prevent the achievement of even better results. The distributional models that include contextual information can bring advantages to this area, but these models are still scarcely explored. Therefore, this work studies the advantages of incorporating syntactic information in the distributional models, fostering for better results in semantic similarity approaches. For that purpose, the current work explore existing lexical and distributional techniques regarding the measurement of word similarity in Brazilian Portuguese. Experiments were carried out with the lexical database WordNet, using different techniques over a standard dataset. The results indicate that word embeddings can cover words out of vocabulary and have better results in comparison with lexical approaches. The main contribution of this article is a new approach to apply syntactic context in the training process of word embeddings to a Brazilian Portuguese corpus. The comparison of this model with the outcome of the previous experiments shows sound results and presents relevant complementary aspects.
When biblicalhumanities.org launched in 2013, we naively assumed that software developers would know how to take advantage of high quality, innovative data if it were published in open formats under free licenses. We understood the importance of building communities that know how to take advantage of this data, software and frameworks that make it easily accessible, and tailoring data to specific use cases, but we assumed that this would happen if we just made the data freely available. Freely licensed base texts, morphologies, contextual glosses, syntax treebanks, discourse analyses, lexicons, textual variants, images, conjectural emendations, and grammars are available now, and some are at least as good as commercial resources, but they are not yet widely used. In the last year or so, Jupyter Notebooks and academic conferences have started to bring this data into academic study, but we have barely begun to tap its potential for translation, language learning, and scripture engagement for those who know the languages. To realize this potential, we need to understand the use cases and mindset of potential users and create software and other resources tailored to their needs. Over time, we need to gather feedback from these same users, walk with them as they learn and grow, and welcome them into communities that will create the next generation of resources.
This contains data files needed for FinMeter. This data is complementary for FinMeter Python library described in: Mika Hämäläinen and Khalid Alnajjar (2019). Let's FACE it. Finnish Poetry Generation with Aesthetics and Framing. In <em>the Proceedings of The 12th International Conference on Natural Language Generation</em>. Sources: The pretrained vectors for Finnish (es - I know) and English (en) are from E. Grave, P. Bojanowski, P. Gupta, A. Joulin, T. Mikolov, <em>Learning Word Vectors for 157 Languages. Creative Commons Attribution-Share-Alike License 3.0</em>. See https://fasttext.cc/docs/en/crawl-vectors.html The word2vec model trained on the Finnish Internet ParseBank is from Kanerva, Jenna; Luotolahti, Juhani; Laippala, Veronika; Ginter, Filip: Syntactic N-gram Collection from a Large-Scale Corpus of Internet Finnish. Proceedings of the Sixth International Conference Baltic HLT. 2014. paper. Creative Commons Attribution-ShareAlike 4.0 International License. See http://bionlp.utu.fi/finnish-internet-parsebank.html The Finnish concreteness data has been automatically translated from Brysbaert, Marc, Amy Beth Warriner, and Victor Kuperman. "Concreteness ratings for 40 thousand generally known English word lemmas." <em>Behavior research methods</em> 46.3 (2014): 904-911. Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License. see http://crr.ugent.be/archives/1330
Emotion science relies upon both facial and scene stimuli that evoke targeted emotions with regard to hedonic valence (i.e., pleasant or unpleasant) and emotional arousal or intensity. While both stimulus types have well established normative ratings via pleasantness and activation/arousal ratings, there are few direct comparisons of these stimuli. Using an auditory probe reaction time paradigm (probe RT), emotional scenes (i.e., the IAPS) were found to be more emotionally engaging compared to facial stimuli (i.e., KDEF). However, some facial sets are more engaging than others--namely the NimStim evokes significantly greater emotional arousal than the KDEF. Here, we plan to compare facial stimuli and scenes from the IAPS using the probe RT paradigm. We plan to compare the two stimulus types (faces vs. scenes) and compare the two facial stimulus sets (KDEF vs. NimStim) with the scenes from the IAPS. We predict that the NimStim will be significantly more emotionally engaging than the KDEF stimuli and that the emotional scenes will be significantly more engaging than both facial stimulus sets. While emotional scenes may be more emotionally engaging than emotional facial stimuli, the current study will extend the literature by demonstrating that certain facial stimuli (e.g., the NimStim) are more emotionally engaging than others (i.e., the KDEF). These findings, for instance, may inform social anxiety research, which relies heavily on facial stimulus sets, by providing a reaction time index of emotional engagement that compliments self-report emotional judgments to guide the emotion researcher in stimulus selection.
espanolEn el presente trabajo se delinea la aportacion de la lengua vasca a la formacion de la norma castellana durante la Edad Media y el Siglo de Oro. Para llegar a tal fin, se perfila primero el marco historico de redes sociales y linguisticas operantes en el contacto del castellano con el euskera desde los origenes de la convivencia vasco-latino-romanica hasta el siglo XVII y se tiene en cuenta, despues, la doble direccion en el contacto vasco-castellano-romanico, sobrevenido historicamente tanto en la direccion del euskera hacia el castellano como del castellano al vascuence, sin olvidar que ha afectado tambien a territorio hoy frances. EnglishThe present article studies the Basque Language contribution to the Castilian linguistic Norm in the Middle Ages and the Golden Age. For this purpose will be first designed the historical frames operating since ancient times on the linguistic contact between Basque and Romance Languages in order to explain how the direction of the Basque-Romance contact has occurred in both directions respectively, taken into account that France domain has been also concerned.
Issues surrounding English for Academic Purposes and its use by non-native English speakers in higher education have become increasingly significant in recent years, fueled both by increased international student mobility and increased linguistic and cultural diversity within and outside of the student body. As well as posing language-related challenges, the transfer of non-native English speakers to an English speaking foreign university also demands the negotiation of new university expectations, channeled through a new cultural environment. While academic literacies research has identified that concepts such as power, identity, and culture play a role in academic writing, the navigation of these aspects in academic writing have not yet been studied thoroughly. Consequently, this study analyzes the ways in which non-native English speaking (NNES) students articulate their navigation of power, identity, and culture within their own academic writing at a tertiary institution in Ireland. Data informing this study was gathered through questionnaires and followed by in-depth case studies of students interview responses analyzed through discourse analysis. The findings suggest that while participants generally positively reflect on their ability to negotiate academic writing through the English language, there is nonetheless a high level of conflict between dominant linguistic norms and the students’ expression of their identity and culture. These findings suggest a need to increase focus on academic literacies in tertiary institutions in order to aid the negotiation of these aspects and to increase the academic success of non-native English speakers.
The purpose of this paper is to find a methodology for parsing Korean by the method of ''Universal Dependencies''. Universal Dependencies expresses the annotation marking system that can be applied to the formal patterns and syntactic relationships of various types of languages. The UD project presents a framework for consistently annotated multi-lingual treebank corpus for various languages around the world through the ''CoNLL shared task'' and is currently applied to more than 50 languages. This flow has a great influence on research related to not only the UD project itself but also the syntactic annotation for individual languages, and efforts are being made to apply the universally dependent syntax analysis marking system in the UD according to the characteristic of the individual language. When annotation labels cannot be applied due to the nature of the Korean language in consideration of compatibility, they need to be excluded. Among DEPREL labels, several labels are difficult to apply in Korean when looking at the specific items. In this paper,, DEPREL overall annotation system of UD was translated and a representative category of elements that needed to be transformed was selected to discuss the actuality of linguistic convergence and its application.
Creating an opinionated lexicon is an important step towards a reliable social media analysis system. In this article we are proposing an approach and describing an experiment to build an Arabic polarised lexical database from analysing online implicitly and explicitly rated customer reviews. These reviews are written in modern standard Arabic and Palestinian/Jordanian dialect. Therefore, the produced lexicon contains casual slangs and dialectic entries used by the online community, which is useful for sentiment analysis of informal social media micro-blogs. We have extracted 28,000 entries from processing 15,100 reviews and by expanding the initial lexicon through Google translate. We calculated an implicit rating for every review driven by its text to address the problem of ambiguous opinions of certain online posts, where the text of the review does not match the given rating (the explicit rating). Each entry was given a polarity tag and a confidence score. High confidence scores have increased the precision of the polarisation process. Explicit rating has increased the coverage and confidence of polarity.
Death is a vague, frightening and abstract concept, mostly considered as a taboo. The current study investigates the ways of conceptualizing death among Persian language speakers in a representative corpus of Persian texts. It seeks answer to the question: What kind of linguistic tools are used by Persian language speakers in order to conceptualize the phenomenon of death and what cultural elements are involved in this regard.To answer the research question, we used the Persian Linguistic Database (PLDB) and a selection of obituaries and epitaphs as data and attempted to identify and extract all the expressions which were directly or indirectly related to the concept of death. Using the conceptual metaphor theory (CMT), by Lakoff and Johnson (1987), and cultural conceptualization by Sharifian (2011), the ways death is conceptualized were identified on the basis of a cognitive-cultural approach.
Treebanks traditionally treat punctuation marks as ordinary words, but linguists have suggested that a tree’s “true” punctuation marks are not observed (Nunberg, 1990). These latent “underlying” marks serve to delimit or separate constituents in the syntax tree. When the tree’s yield is rendered as a written sentence, a string rewriting mechanism transduces the underlying marks into “surface” marks, which are part of the observed (surface) string but should not be regarded as part of the tree. We formalize this idea in a generative model of punctuation that admits efficient dynamic programming. We train it without observing the underlying marks, by locally maximizing the incomplete data likelihood (similarly to the EM algorithm). When we use the trained model to reconstruct the tree’s underlying punctuation, the results appear plausible across 5 languages, and in particular are consistent with Nunberg’s analysis of English. We show that our generative model can be used to beat baselines on punctuation restoration. Also, our reconstruction of a sentence’s underlying punctuation lets us appropriately render the surface punctuation (via our trained underlying-to-surface mechanism) when we syntactically transform the sentence.
Abstract: Temperatures above 20° Celsius have shown to adversely impact human behavior, leading to increased aggression and violence. Climate change will contribute to both the magnitude and severity of this pattern as temperatures continue their rise. Contributions to this field of research have only recently begun to analyze online behavior and language as a proxy for hedonic state, or well-being. From a development perspective this study is relevant since the poor tend to live in some of the warmest regions on earth, and would thus be disproportionately impacted by increased temperatures. We use several sources of data; U.S. based daily statewide temperature data from 2016 through 2017, as well as localized viewer chat data from a live video streaming website. We will sort chatting comments looking for key words (i.e. hate speech, swearing, etc.), and with the use of a word rating system we then assess the overall mood of the chatters contingent on high temperature readings on the precise day of the communications. After controlling for spatiotemporal fixed effects, we find strong evidence that hedonic state decreases above 20°c.
Abstract The topic of this paper is the interaction of aspectual verb coding, information content and lengths of verbs, as generally stated in Shannon’s source coding theorem on the interaction between the coding and length of a message. We hypothesize that, based on this interaction, lengths of aspectual verb forms can be predicted from both their aspectual coding and their information. The point of departure is the assumption that each verb has a default aspectual value and that this value can be estimated based on frequency – which has, according to Zipf’s law, a negative correlation with length. Employing a linear mixed-effects model fitted with a random effect for LEMMA, effects of the predictors’ DEFAULT – i.e. the default aspect value of verbs, the Zipfian predictor FREQUENCY and the entropy-based predictor AVERAGE INFORMATION CONTENT – are compared with average aspectual verb form lengths. Data resources are 18 UD treebanks. Significantly differing impacts of the predictors on verb lengths across our test set of languages have come to light and, in addition, the hypothesis of coding asymmetry does not turn out to be true for all languages in focus.
Antiaddictive social advertising is a special speech genre of modern communication with specific features determined by the target setting and the chosen strategy. Advertising can be considered as a special functional style, within which separate genres are distinguished, first of all commercial advertising and social advertising. Social advertising, which refers to ethical categories, has become especially popular because society is faced with such problems, the solution of which depends on mass behavior. Advertising has become an integral part of the daily life of a person. Since advertising is mass-replicated in the media, it can enter the consciousness of the addressee, even against his/her will and desire, no wonder advertising is sometimes defined as the “fifth power” after media, whose power is considered the “fourth power”. Therefore, it is so important for advertising to follow the moral principles of society, it is so important to comply with modern ethical and linguistic norms. If commercial advertising is widespread, the antiaddictive one is little known. Many specialists in advertising argue for the need of the strategy of “shock” advertising, a remarkable feature of which is hyperbolization, used to cause fear. In antiaddictive advertising there is often a morphological imperative in the meaning of categorical motivation. It is justified as such advertizing has to be as much as possible appellate, has to be understood unambiguously. Anti-drug social advertising should show positive, motivation to a healthy lifestyle.
In two experiments, the influence of inducing negative mood on cognitive performance was explored by analyzing physical arm reaching movements as indicators of mind wandering. Mood was induced by viewing a series of six photos per mood condition that were previously established for their emotionally valenced and arousal ratings. A reach tracking device recorded three metrics of arm movement that were expected to reflect instances of mind wandering: initiation latency, movement time, and arm curvature. In the first experiment, 29 participants were randomly assigned into one of two induced-mood groups, negative mood (n = 15) or neutral mood (n = 14). Participants performed a simple Go/No-go task in which arm movements were detected by the reach tracker. The first experiment indicated that the mood inducement was successful but the effect of negative mood on either self-reported mind wandering or variances in arm movement were not significant. Thus, the second experiment prompted the change to a visual-search target-selection task in which variances in initiation latency, movement time, and curvature were expected to be more pronounced. The second experiment consisted of 23 participants who were also randomly assigned to either negative (n = 12) or neutral (n = 11) mood condition. The second experiment revealed that the mood induction was still successful but that there were still no significant effects observed between mood and indicators of mind wandering. Though the results of this study did not reflect initial predictions, it may suggest that low-arousing negative moods in healthy individuals are not associated with increased mind wandering.
Abstract This article explores the continuing linguistic impact of the Mandarin Union Version by investigating and contrasting two Chinese translations of William Paul Young's global bestseller The Shack (2007): the Traditional Chinese version Xiaowu (《小屋》, 2009) and the Simplified Chinese version Pengwu (《棚屋》, 2010). Ever since its publication, the Mandarin Union Version has served as the predominant Bible within Mandarin-speaking Protestant communities across the world. This has brought about the standardisation of terminology in Chinese Protestantism. The Shack, though widely marked as a Christian novel, is also known for its unconventional fictional representations of Christianity that some Christians think depart from orthodoxy. Both Xiaowu and Pengwu were published by non-Christian publishing houses for a general readership. However, Xiaowu, translated by a Christian, exhibits a significant number of phrases that specifically belong to Chinese Christian terminology shaped by the Mandarin Union Version. Pengwu is a contrast in this regard. By comparing extracts from these two Chinese versions, this article highlights how far the Mandarin Union Version has contributed to the formation of the linguistic repertoire of Mandarin-speaking Christian translators as well as linguistic norms for translated Christian-themed texts into Chinese.
This study investigated the associations of imageability with fear reactivity. Imageability ratings of four word classes: positive and negative (i) emotional and (ii) propriosensitive, neutral and negative (iii) theoretical and (iv) neutral concrete filler, and fear reactivity scores – degree of fearfulness towards different situations (TF score) and total number of extreme fears and phobias (EF score), were obtained from 171 participants. Correlations between imageability, TF and EF scores were tested to analyze how word categories and their valence were associated with fear reactivity. Imageability ratings were submitted to recursive partitioning. Participants with high TF and EF scores had higher imageability for negative emotional and negative theoretical words. The correlations between imageability of negative emotional words and negative theoretical words for EF score were significant. Males showed stronger correlations for imageability of negative emotional words for EF and TF scores. High imageability for positive emotional words was associated with lower fear reactivity in females. These findings were discussed with regard to negative attentional bias theory of anxiety, influence on emotional systems, and gender-specific coping styles. This study provides insight into cognitive functions involved in mental imagery, semantic competence for mental imagery in relation to fear reactivity, and a potential psycholinguistic instrument assessing fear tendency.
We introduce a novel transition system for discontinuous constituency\nparsing. Instead of storing subtrees in a stack --i.e. a data structure with\nlinear-time sequential access-- the proposed system uses a set of parsing\nitems, with constant-time random access. This change makes it possible to\nconstruct any discontinuous constituency tree in exactly $4n - 2$ transitions\nfor a sentence of length $n$. At each parsing step, the parser considers every\nitem in the set to be combined with a focus item and to construct a new\nconstituent in a bottom-up fashion. The parsing strategy is based on the\nassumption that most syntactic structures can be parsed incrementally and that\nthe set --the memory of the parser-- remains reasonably small on average.\nMoreover, we introduce a provably correct dynamic oracle for the new transition\nsystem, and present the first experiments in discontinuous constituency parsing\nusing a dynamic oracle. Our parser obtains state-of-the-art results on three\nEnglish and German discontinuous treebanks.\n
Several studies have reported specific semantic dissociations in brain damaged patients. These semantic dissociations concern particularly the impairment of one domain of knowledge (living concepts) whilst the other is relatively spared (nonliving concepts). The origin of these dissociations remain controversial. A important number of semantic theories have advanced different explanations for this phenomenon. Among others we include: the sensory-functional theory (Warrington & McCarthy, 1983), the organized content hypothesis, the domain specific hypothesis (Caramazza & Shelton, 1998) and the conceptual structure account (Moss, Tyler, & Devlin, 2002). However, at first time some authors have suggested that category-specific semantic deficit for living things may be due to a lower familiarity and frequency and a more visual complexity of these concepts compared with nonliving concepts (Funnell & Sheridan, 1992). In addition, recently researches suggest that familiarity varies according to the gender and therefore it may to influence the emergence of category-specific semantic deficits. Thus, Laiacona, Barbarotto and Capitani (1998) showed that females with Alzheimer's disease were more impaired with nonliving categories, whereas males patients showed more deficit with living categories. Likewise, studies from normal subjects have shown a gender-category interaction in a semantic fluency task, females performed better with fruits and males with tools (Capitani, Laiacona, & Barbarotto, 1999). The goals of this study are a) to evaluate the familiarity of concepts according to domains (living and nonliving) and categories and b) to explore the gender effect on familiarity ratings.
ABSTRACT Issues surrounding English for Academic Purposes (EAP) and its use by English as an additional language (EAL) students in higher education have become increasingly significant in recent years, fueled both by increased international student mobility and increased linguistic and cultural diversity within and outside of the student body. As well as posing language-related challenges, the transfer of EAL students to an English-speaking foreign university also demands the negotiation of new university expectations, channeled through a new cultural environment. While Academic Literacies research has identified that concepts such as power, identity, and culture play a role in academic writing, students’ own perceptions remain relatively unexplored. Consequently, this study analyzes the ways in which EAL students articulate their relationship with academic writing at a tertiary institution in Ireland. Data for this study were gathered through questionnaires and interviews and analyzed through discourse analysis through a critical lens. The findings suggest that while participants generally positively reflect on their ability to negotiate academic writing through the English language, there is nonetheless a high level of conflict between dominant linguistic norms and the students’ expression of their identity and culture.
Mirror-sensory synesthetes mirror the pain or touch that they observe in other people on their own bodies. This type of synesthesia has been associated with enhanced empathy. We investigated whether the enhanced empathy of people with mirror-sensory synesthesia influences experience of situations involving touch or pain, and whether it affects their prosocial decision making. Mirror-sensory synesthetes (N=18, all female), verified with a touch-interference paradigm, were compared to a similar number of age-matched control individuals (all female). Participants viewed arousing images depicting pain or touch; we recorded subjective valence and arousal ratings, and physiological responses, hypothesizing more extreme reactions in synesthetes. The subjective impact of positive and negative images was stronger in synesthetes than in control participants; the stronger the reported synesthesia, the more extreme the picture ratings. However, there was no evidence for differential physiological or hormonal responses to arousing pictures. Prosocial decision making was assessed with an economic game assessing altruism, in which participants had to divide money between themselves and a second player. Mirror-sensory synesthetes donated more money than non-synesthetes, showing enhanced prosocial behaviour, and also scored higher on the Interpersonal Reactivity Index as a measure of empathy. Our study demonstrates the subjective impact of mirror-sensory synesthesia and its stimulating influence on prosocial behaviour.
We explore whether it is possible to leverage eye-tracking data in an RNN dependency parser (for English) when such information is only available during training, i.e., no aggregated or token-level gaze features are used at inference time. To do so, we train a multitask learning model that parses sentences as sequence labeling and leverages gaze features as auxiliary tasks. Our method also learns to train from disjoint datasets, i.e. it can be used to test whether already collected gaze features are useful to improve the performance on new non-gazed annotated treebanks. Accuracy gains are modest but positive, showing the feasibility of the approach. It can serve as a first step towards architectures that can better leverage eye-tracking data or other complementary information available only for training sentences, possibly leading to improvements in syntactic parsing.
Abstract This chapter describes the data structure of the Rhapsodie Treebank and discusses methodological issues stemming from the complexity of this structure, articulated around three independent, non-aligned, hierarchies: Microsyntactic, macrosyntactic and prosodic, and the challenging questions to be resolved in this context. It discusses the specific problems posed by the simultaneous processing of the phonological stream (prosodic level) and the orthographic stream (syntactic level), which are often far from being isomorphic in French, and the related problem of the processing of disfluent and/or overlapped strings, which have not the same representation in the syntactic and the prosodic hierarchy. Then, it presents the formats adopted to encode prosodic and syntactic annotations and query them simultaneously, given that the prosodic architecture is a non-recursive time-aligned representation while the syntactic one is a recursive tree-based representation.
This chapter provides a history of 'the King's English' as a context for an analysis of language, history and power in The Merry Wives of Windsor and the second tetralogy. The trope is used as a rhetorical and ideological tool in performatives. Associated with temperance and honesty, 'the King's English' belongs to a set of defining values of true Englishness. The project to produce this linguistic norm coincides with a homologous project to produce a stable, monetary system of 'good' coin through exclusion of 'bad', 'counterfeit' or 'clipped' coin. These projects testify to a shift of the centre of economic and cultural gravity from the court to the merchant citizen class. Shakespeare's one English comedy centred on English citizens which features his one use of 'the King's English' is shown to engage critically with this ideology, and to set against it an idea of 'our English' as an inclusive mix, the 'gallimaufry' loved by the linguistically extravagant gentleman John Falstaff. The comedy draws out the implications of the banishment of Falstaff in the second tetralogy, which sets history against the project of cultural reformation ideology to produce (the) 'true' English.
This paper examines the relation between gendered language and the processing of non-stereotypical gender representation through a psycholinguistic priming experiment consisting of a self-paced reading test. The experiment tests two things: The processing ease of 3rd person singular pronouns that either match or mismatch the stereotypical gender of their referents; and whether sentences with gendered language affect this. Processing ease is measured by reading time. Sentences with gendered language are used as priming; and pronouns with matching or mismatching stereotypicality as targets. The results showed no difference in the reading time of matching and mismatching pronouns, and no priming effect was found. This could point to the fact that gender stereotypical mismatch does not affect the informants; and that gendered language is so well-integrated in our language that it cannot prime for gender stereotypes. Moreover, the pronoun han was read significantly faster than the pronoun hun, which could indicate that the masculine is expected as a linguistic norm. This is considered in relation to the reflections on the missing priming effect as a result of gendered language being the norm.
This paper focuses on the process and principle of parsing, which is an essential task for machine to understand the syntactic, semantic structure of a sentence. First, a series of machine analysis procedures such as word segmentation, part-of-speech tagging and parsing of Chinese sentences are visually represented by using Cparser, a rule-based constituency parser developed by Peking University. Next, to better understand parsing mechanism, we explain in detail how the linguistic knowledge is embodied in the lexical, syntactic and semantic component of Cparser, showing their complex interplay that allows automatic parsing. As a practical example, a Chinese textbook treebank is also constructed using Cparser. According to the theoretical and practical discussion in this paper, Peking University Cparser, which is easy to reflect and modify linguistic knowledge, is expected to be widely used as an analysis and verification tool for Chinese grammar research.
This paper presents the first gold-standard resource for Russian annotated with compositionality information of noun compounds. The compound phrases are collected from the Universal Dependency treebanks according to part of speech patterns, such as ADJ+NOUN or NOUN+NOUN, using the gold-standard annotations. Each compound phrase is annotated by two experts and a moderator according to the following schema: the phrase can be either compositional, non-compositional, or ambiguous (i.e., depending on the context it can be interpreted both as compositional or noncompositional). We conduct an experimental evaluation of models and methods for predicting compositionality of noun compounds in unsupervised and supervised setups. We show that methods from previous work evaluated on the proposed Russian-language resource achieve the performance comparable with results on English corpora.
The article refers to the concept of intelligentsia as a social group which exerts significant influence on Polish standard patterns. Although the term intelligentsia is vague and questionable, it is well-established term in Polish linguistics, especially in sociolinguistics. Author argues that science communicators (young professional researchers, science journalists, PhD students) represent the young intelligentsia, because these well-educated people pursue their intellectual development and they have sense of public duty. The article examines standard of popular science texts in Internet, new tendencies in written Polish and attitude of young intelligentsia toward traditional linguistic norm. The errors (esp. punctuation and syntax) exemplify impact of technological changes and phenomenon of secondary orality. It would be useful for science communicators to edit carefully their texts. Both researchers and journalists need to improve their writing skills permanently. Nevertheless it must be emphasized that school education and competent teachers seem to have important influence on the linguistic patterns.
The communicative role of nonlinear vocal phenomena remains poorly understood since they are difficult to manipulate or even measure with conventional tools. In this study parametric voice synthesis was employed to add pitch jumps, subharmonics/sidebands, and chaos to synthetic human nonverbal vocalizations. In Experiment 1 (86 participants, 144 sounds), chaos was associated with lower valence, and subharmonics with higher dominance. Arousal ratings were not noticeably affected by any nonlinear effects, except for a marginal effect of subharmonics. These findings were extended in Experiment 2 (83 participants, 212 sounds) using ratings on discrete emotions. Listeners associated pitch jumps, subharmonics, and especially chaos with aversive states such as fear and pain. The effects of manipulations in both experiments were particularly strong for ambiguous vocalizations, such as moans and gasps, and could not be explained by a non-specific measure of spectral noise (harmonics-to-noise ratio) – that is, they would be missed by a conventional acoustic analysis. In conclusion, listeners interpret nonlinear vocal phenomena quite flexibly, depending on their type and the kind of vocalization in which they occur. These results showcase the utility of parametric voice synthesis and highlight the need for a more fine-grained analysis of voice quality in acoustic research.
Abstract This article presents results from a study on hybrid linguistic norms in translated articles from New York Times made available on UOL website. According to Faraco (2008) and Bagno (2012), there is a difference between norma padrão (a prescriptive norm, but not based on usage) and norma culta (an alternative, usage-based norm). The first one combines normative rules that determine correct linguistic forms, but generally hard to follow by most users, while the second one brings together a set of linguistic forms frequently employed by users, because they are more intuitively accessible, although not subscribed by the conservative standard norm (norma padrão) commonly taught in grammar books and in writing style manuals. The research was meant to verify if journalistic texts translated from English have been as permeable to linguistic forms not subscribed by the prescriptive standard norm, as the ones originally written in Portuguese have proven to be.
Location: Dewberry Hall With over 34,000 students representing 123 countries, George Mason University is a vastly diverse university with students bringing different learning experiences and skill sets with them into the classroom. The study that our team conducted analyzes the essays of native English speakers, as well as the essays of students for whom English is their second language. Our objective when conducting this research was to observe the essays for signs of syntactic complexity and patterns of language errors. Specifically, we looked for subordinating clauses, transitions, subject/verb agreement, article usage, run-on sentences, and fragments. We found that some errors in L1 and L2 populations were consistent with our expectations, but others reveled a more complex understanding of the linguistic norms of the groups studied. The results of these findings will give professors of all disciplines and modalities insight to the challenges that first-year L1 and L2 students confront when faced with a writing assignment.
Contextualized embeddings, which capture appropriate word meaning depending\non context, have recently been proposed. We evaluate two meth ods for\nprecomputing such embeddings, BERT and Flair, on four Czech text processing\ntasks: part-of-speech (POS) tagging, lemmatization, dependency pars ing and\nnamed entity recognition (NER). The first three tasks, POS tagging,\nlemmatization and dependency parsing, are evaluated on two corpora: the Prague\nDependency Treebank 3.5 and the Universal Dependencies 2.3. The named entity\nrecognition (NER) is evaluated on the Czech Named Entity Corpus 1.1 and 2.0. We\nreport state-of-the-art results for the above mentioned tasks and corpora.\n
This article proposes a character-level neural language model (NLM) that is based on quantum theory. The input of the model is the character-level coding represented by the quantum semantic space model. Our model integrates a convolutional neural network (CNN) that is based on network-in-network (NIN). We assessed the effectiveness of our model through extensive experiments based on the English-language Penn Treebank dataset. The experiments results confirm that the quantum semantic inputs work well for the language models. For example, the PPL of our model is 10%–30% less than the states of the arts, while it keeps the relatively smaller number of parameters (i.e., 6 m).
Abstract This chapter is devoted to the presentation of the tools and methods used for the different steps of the semi-automatic syntactic annotation: automatic preprocessing; microsyntactic parsing with the FRMG tool, correction of the parsing with the Arborator tool, agreement analysis, post-validation correction, and development of the final format of the Rhapsodie syntactic treebank. As FRMG is a parser for written French that was not configured to analyze disfluencies and reformulation, we used our manual pile marking to unfold the piles and produce a series of simplified “sentences” with only government relations. Despite having two annotators plus a validator for the corrections, we found a substantial number of errors in the post-validation procedure by using a set of rules to determine the well-formedness of the trees.
This talk will explore the role of individual social actors and their communities and networks in the formation, maintenance and dissolution of linguistic norms. It will consider several standardisation episodes in the history of Old and Early Middle English and connect them to other unification processes in the political and cultural history of England at the time. It will be suggested that the suppression of variability on the linguistic level often accompanies, or is a symptom of, a similar suppression on the ideological level. Political, religious, legal, and linguistic processes mingle in various ways in this period (as they do today) to replicate and enhance the social order, to support a reform movement, or to refute dissent. \nThree case studies will offer insights into these processes: 1) shire courts and the ‘standardised’ lexis of the Anglo-Saxon Chronicle in the reign of King Alfred and Edward the Elder; 2) chancery norms and charters of the eleventh century; and 3) religious reform and linguistic focusing in the thirteenth century.
This is a work-in-progress report, which aims to share preliminary results of a novel sequence-to-sequence schema for dependency parsing that relies on a combination of a BiLSTM and two Pointer Networks (Vinyals et al., 2015), in which the final softmax function has been replaced with the logistic regression. The two pointer networks co-operate to develop a latent syntactic knowledge, by learning the lexical properties of "selection" and the lexical properties of "selectability", respectively. At the moment and without fine-tuning, the parser implementation gets a UAS of 93.14% on the English Penn-treebank (Marcus et al., 1993) annotated with Stanford Dependencies: 2-3% under the SOTA but yet attractive as a baseline of the approach.