Papers reviewed and determined not to be word norm studies. Use the flag icon to report errors or suggest re-inclusion.
18625 papers
Understanding the sense of discourse relations between segments of text is essential to truly comprehend any natural language text. Several automated approaches have been suggested, but all rely on external resources, linguistic feature engineering, and their processing pipelines are built from substantially different models. In this paper, we introduce a novel system for sense classification of shallow discourse relations (FR system) based on focused recurrent neural networks (RNNs). In contrast to existing systems, FR system consists of a single end-to-end trainable model for handling all types and senses of discourse relations, requires no feature engineering or external resources, is language-independent, and can be applied at the word and even character levels. At its core, we present our novel generalization of the focused RNNs layer, the first multi-dimensional RNN-attention mechanism for constructing text/argument embeddings. The filtering/gating RNN enables downstream RNNs to)
The estimated period in which human colonization of Madagascar began has expanded recently to 5000–1000 y B.P., six times its range in 1990, prompting revised thinking about early migration sources, routes, maritime capability and environmental changes. Cited evidence of colonization age includes anthropogenic palaeoecological data 2500–2000 y B.P., megafaunal butchery marks 4200–1900 y B.P. and OSL dating to 4400 y B.P. of the Lakaton’i Anja occupation site. Using large samples of newly-excavated bone from sites in which megafaunal butchery was earlier dated >2000 y B.P. we find no butchery marks until ~1200 y B.P., with associated sedimentary and palynological data of initial human impact about the same time. Close analysis of the Lakaton’i Anja chronology suggests the site dates <1500 y B.P. Diverse evidence from bone damage, palaeoecology, genomic and linguistic history, archaeology, introduced biota and seafaring capability indicate initial human colonization of Madagascar 1350–1)
Language in its highest complexity is a unique human faculty with simultaneous translation being among the most demanding language task involving both linguistic and executive functions. In this context, bilingually grown up individuals as well as simultaneous interpreters (SIs) represent appropriate groups for studying expertise-related neural adaptations in the human brain. The present study was performed to examine if a domain-specific neural network activation pattern, constituted by brain regions involved in speech processing as well as cognitive control mechanisms can be detected during a task-free resting state condition. To investigate this, electroencephalographic (EEG) data were recorded from 16 SIs and 16 age and gender-matched multilingual control subjects. Graph-theoretical network analyses revealed interhemispheric hyperconnectivity between the ventral part of the prefrontal cortex (pars opercularis and pars triangularis) and the dorsolateral prefrontal cortex (DLPFC) in)
This study uses a maze navigation task in conjunction with a quasi-scripted, prosodically controlled speech task to examine acoustic and articulatory accommodation in pairs of interacting speakers. The experiment uses a dual electromagnetic articulography set-up to collect synchronized acoustic and articulatory kinematic data from two facing speakers simultaneously. We measure the members of a dyad individually before they interact, while they are interacting in a cooperative task, and again individually after they interact. The design is ideally suited to measure speech convergence, divergence, and persistence effects during and after speaker interaction. This study specifically examines how convergence and divergence effects during a dyadic interaction may be related to prosodically salient positions, such as preceding a phrase boundary. The findings of accommodation in fine-grained prosodic measures illuminate our understanding of how the realization of linguistic phrasal structure)
Cilin is one of the most popular semantic knowledge bases in Chinese information processing. Due to its coding scheme and taxonomical arrangement, some semantic relations among words are not explicitly shown. Our work aims to characterize its semantic relations by adding tags and compound codes to optimize the taxonomy and hierarchy of Cilin. Experiment results show that using the tag-augmented Cilin as a knowledge base improves the performance in the task of semantic similarity measurement.
This study examines how the acoustic input (the surface form) and the abstract linguistic representation (the underlying representation) interact during spoken word recognition by investigating left-dominant tone sandhi, a tonal alternation in which the underlying tone of the first syllable spreads to the sandhi domain. We conducted an auditory-auditory priming lexical decision experiment on Shanghai left-dominant sandhi words, in which each disyllabic target ([tɕi55 dɛ31] “egg”) was preceded by monosyllabic primes either sharing the same underlying tone ([tɕi55]), surface tone ([tɕi53] “machine”), or being unrelated to the tone of the first syllable of the sandhi targets ([tɕi24] “to remember”). Results showed a surface priming effect, but not an underlying priming effect. Moreover, the surface priming did not interact with speakers’ familiarity ratings to the sandhi targets. The results are discussed in the context of how phonological opacity, productivity, and the directionality of tone sandhi patterns influence the representation of tone sandhi words as well as how the lexicality of the primes and the participants’ usage pattern of Shanghai may have influenced the results.
Many recent studies, academic and non-academic alike, have argued that the use of Urdu in Bollywood has started to decline. These studies, important as they are, however, suffer from some limitations. They are either impressionistic or based on non-representative data. Furthermore, they do not specify the object of the study or the site of the assumed decline of Urdu. Therefore, it remains vague which element of Urdu, for example sounds, words, syntax or script is under investigation. Similarly, it is not clear which component of film for example titles, dialogues or songs are experiencing the decline. Fulfilling this research gap this paper makes two contributions. Analyzing songs from 1959 to 2010’s, it empirically demonstrates the decline by documenting the shift in the pronunciation of the sounds /kh̲/, /gh̲/, and /q/ from the Urdu to Hindi phonetic norms. Singers from the 1990’s, unlike those from the previous generations, merge them with the sounds /kh/, /gh/, and /q/. The paper also makes a methodological contribution in that it shows how language in cinema can be studied empirically using a corpus.
We conducted a semantic similarity study of semantic concepts in the context of the Holy Book Quran. Semantic similarity examines the degree of likeness and shared common properties of two concepts. For example, the Quranic concept of Allah and God will result in a high score of semantic similarity, whereas hell and paradise will yield in a low score because of its extremely different attributes and semantic features. Apart from that, we also delivered the Quranic concept semantic similarity standard dataset which consists of some pairs of Quranic concept along with its similarity score, which was manually annotated by human raters. This dataset resulted in the score of inter-annotator agreement 0.63, not far from the the ones yielded by some well-known datasets such as WordSim and Simlex. Furthermore, to measure the semantic similarity score, we chose the knowledge-based approach by utilizing lexical database properties such as the length and depth of a synonym set (synset). We then applied it to Yuhua Li equation, which has been considered to be the baseline among researchers within the problem of semantic similarity. In terms of the result, our system gained Pearson's correlation 0.33 and Spearman's 0.19. By considering inter-annotator agreement 0.63 that our Quranic standard dataset has as the upper bound score, there are still quite large room for improvement to better mimicking Muslim's intuition to measure the degree of similarity of concepts within the domain of Quran.
The Universal Dependencies project is currently comprised of 71 languages and 122 treebanks, and aims to find morphological and syntactic characteristics that can be applied to multiple languages for parallel language processing. In this paper, we introduce Universal POS, which is a morphological tagset for UD, and propose a method to automatically convert existing Korean morphological tagset into UPOS. In order to apply the UPOS tagset, which is based on refraction words such as English, to the Korean language, it is necessary to try a one-to-many mapping between the UPOS individual tag and the 21st century Sejong tag combination. (Yonsei University)
Because the most common transition systems are projective, training a transition-based dependency parser often implies to either ignore or rewrite the non-projective training examples, which has anadverse impact on accuracy. In this work, we propose a simple modification of dynamic oracles, which enables the use of non-projective data when training projective parsers. Evaluation on 73~treebanks shows that our method achieves significant gains (+2 to +7 UAS for the most non-projective languages) and consistently outperforms traditional projectivization and pseudo-projectivizationapproaches.
In today’s increasingly divided political climate there is a need for a tool that can compare news articles and organizations so that a user can receive a wider range of views and philosophies. NewsAnalyticalToolkit allows a user to compare news sites and their political articles by coverage, mood, sentiment, and objectivity. The user can sort through the news by topic, which was determined using Natural Language Processing (NLP) and Latent Dirichlet Allocation (LDA). LDA is a probabilistic method used to discover latent topics within a series of documents and cluster them accordingly. Each news article can be considered a mix of multiple topics and LDA assigns a set of topics to each with a probability of it pertaining to that topic. For each topic, a user can then discover the coverage, mood, sentiment and objectivity expressed by each author and site. The mood was determined using IBM Watsons ToneAnalyzerV3, which uses linguistic analysis to detect emotional, social and language tones in written text. The analyzer is based on the theory of psycholinguistics, a field of research that explores the relationship between linguistic behavior and psychological theories. The sentiment and objectivity scores were determined using SentiWordNet, which is a lexical database that groups English words into sets of synonyms and assigns sentiment scores to them. The features were combined to plot an interactive graph of how opinionated versus how analytical an article is, so that the user can click through them to get a better understanding of the topic in question.
Polycentric Spanish Norm Towards the Polish‑Spanish Legal Translation The Spanish, being the official language of Spain and many other countries, is characterized by an important dialectal diversity that is reflected in the differences at all linguistic levels: phonetic, morphological, syntactic and lexico‑semantic, etc. All these differences raise controversies and discussions about the existence of a linguistic norm depending on the perspective that can have a monocentric or polycentric character. In this contribution we present some arguments for the second one. To this end, we rely on translations, starting simultaneously from the semasiological and onomasiological perspective, of some Polish‑Spanish legal terms in which it is essential to take into account, the diatopic variation as well as the norm whose character is polycentric.
This era, in which we currently stand, is an era of public opinion and mass information. People from all around the globe are joined together through various information junctions to create a global community, where one thing from the far east reaches to the people of the far west within seconds. Nothing is hidden, everything and anything can be scrutinized to its core and through these global criticisms and mass discussions of gigantic magnitude, we have reached to the pinnacle of correct decisions and better choices. These pseudo social groups and data junctions have bombarded our society so much that they now hold the forelock of our opinions and sentiments, ergo, we reach out to these groups to achieve a better outcome. But, all this enormous data and all these opinions cannot be researched by a single person, hence, comes the need of sentiment analysis. In this paper we’ll try to accomplish this by creating a system that will enable us to fetch tweets from twitter and use those tweets against a lexical database which will create a training set and then compare it with the pre-fetched tweets. Through this we will be able to assign a polarity to all the tweets by means of which we can address them as negative, positive or neutral and this is the very foundation of sentiment analysis, so subtle yet so magnificent.
The short note describes the chart parser for multimodal type-logical grammars which has been developed in conjunction with the type-logical treebank for French. The chart parser presents an incomplete but fast implementation of proof search for multimodal type-logical grammars using the "deductive parsing" framework. Proofs found can be transformed to natural deduction proofs.
The article deals with the issue of translation as an important means of communication between individuals who speak different languages and belong to different cultures.The article analyzes the translation as interlingual communicative phenomenon.The translation process is determined by the linguistic norms, communicative situations, functional parameters of the original text and translation norms.The role and tasks of an interpreter in the process of intercultural communication of individuals are defined.
This chapter discusses the new and changing conditions for linguistic norms in literary fiction of the post-Soviet era. In particular, it looks at how the interrelationship between the language of literature (<italic>iazyk literatury</italic>) and the standard language (<italic>literaturnyi iazyk</italic>) has been challenged by several processes of sociolinguistic change, including initiatives in language policy.
The article discusses the terms that nominate the language of written artifacts documented by the Cyrillic on the Ukrainian-Byelorussian lands in the XIV-XVI centuries; the expediency of using the notion “literary language” as to the Ukrainian literary written tradition of the XIV–XVI centuries is clarified; the content of the term “linguistic norm” is outlined, its characteristics in the investigated period are determined.
The linguistic database is also positioned as an actual way of formalizing and organizing phraseological units, terms for designating types of phraseological units. The main principle of systematization of the latter in the study is the thesaurus principle, that is the filling of the paradigm «terminological system – terminological microsystem – terminological subsystem – term», represented by a linguistic database.
Abstract The aim of the contribution is to introduce a database of linguistic forms and their functions built with the use of the multi-layer annotated corpora of Czech, the Prague Dependency Treebanks. The purpose of the Prague Database of Forms and Functions (ForFun) is to help the linguists to study the form-function relation, which we assume to be one of the principal tasks of both theoretical linguistics and natural language processing. We demonstrate possibilities of the exploitation of the ForFun database. This article is largely based on a paper presented at the 16th International Workshop on Treebanks and Linguistic Theories in Prague (Bejček et al., 2017).
Functional alterations of the default mode network (DMN) are frequently reported in psychotic disorders, but the functional role of these alterations remains poorly known. In addition to previous studies that have applied different types of tasks or recorded resting-state neuroimaging data, there has recently been more interest in the use of movie stimuli in studying brain functioning in patient populations, because this could provide a more naturalistic account of brain functioning in real life-like situations. Seventy-one first-episode psychosis (FEP) patients (mean age = 26.0 yrs, 47 (66%) males) and 57 controls (mean age = 26.86 yrs, 24 (42%) males) from the Helsinki Early Psychosis Study watched scenes from the movie Alice in Wonderland (Tim Burton, 2010) during 3 T fMRI-BOLD imaging. We used intersubject correlation (ISC) analysis, in which the correlation between voxel-wise BOLD time series in every within-group pair of subjects is calculated. In this study, time-windowed ISC was calculated with a 10-TR (time of repetition, 1.8 s) window with 1-TR steps over the fMRI time series. In each ISC window, a two-sample t test was performed to obtain a t-statistic time series of differences between the groups. An independent group of control subjects (n = 17, 10 males, mean age 26.5 yrs) rated how emotionally arousing the currently seen events of the stimulus are, producing a time-varying rating used as a regressor. General linear model was used to identify brain regions where the t-statistic time series covaries with the arousal rating. To make the interpretation of results less ambiguous, the arousal rating was divided into high and low arousal regressor by z scoring the rating and taking only the positive and negative values, respectively. Nonparametric clusterwise permutation test was used for statistical inference (cluster-defining threshold of p = 0.05, familywise error corrected threshold of p = 0.05, number of permutations = 5000). Furthermore, by using an experience-sampling setup during the same brain-scanning session, a partially overlapping sample of participants reported how emotionally aroused they were feeling during scanning. The results show significant correlation between the t-statistic time series and low arousal regressor, especially in the DMN including the anterior and posterior cingulate cortex, medial prefrontal cortex, precuneus, and bilateral lateral temporoparietal regions. Closer inspection reveals that during moments of low arousal in the movie stimulus, the ISC of healthy controls goes up but the ISC of patients does not. In the experience-sampling portion of the study, the patients reported more arousal than the control subjects. Intersubject correlation in the DMN depended differentially on arousal in FEP patients and control subjects. More specifically, during moments when the stimulus was rated less emotionally arousing, control subjects’ DMN functioning synchronized more while the patients’ did not. In connection with the difference in reported arousal during the same imaging session, our findings provide preliminary evidence for a contribution of arousal on the functional alterations of the DMN and suggest that this may be related to higher baseline arousal in the patients. Higher arousal and the related distortion of high order integrative functioning that characterizes DMN could contribute to the pathogenesis of psychosis.
Temporal models based on recurrent neural networks have proven to be quite\npowerful in a wide variety of applications. However, training these models\noften relies on back-propagation through time, which entails unfolding the\nnetwork over many time steps, making the process of conducting credit\nassignment considerably more challenging. Furthermore, the nature of\nback-propagation itself does not permit the use of non-differentiable\nactivation functions and is inherently sequential, making parallelization of\nthe underlying training process difficult. Here, we propose the Parallel\nTemporal Neural Coding Network (P-TNCN), a biologically inspired model trained\nby the learning algorithm we call Local Representation Alignment. It aims to\nresolve the difficulties and problems that plague recurrent networks trained by\nback-propagation through time. The architecture requires neither unrolling in\ntime nor the derivatives of its internal activation functions. We compare our\nmodel and learning procedure to other back-propagation through time\nalternatives (which also tend to be computationally expensive), including\nreal-time recurrent learning, echo state networks, and unbiased online\nrecurrent optimization. We show that it outperforms these on sequence modeling\nbenchmarks such as Bouncing MNIST, a new benchmark we denote as Bouncing\nNotMNIST, and Penn Treebank. Notably, our approach can in some instances\noutperform full back-propagation through time as well as variants such as\nsparse attentive back-tracking. Significantly, the hidden unit correction phase\nof P-TNCN allows it to adapt to new datasets even if its synaptic weights are\nheld fixed (zero-shot adaptation) and facilitates retention of prior generative\nknowledge when faced with a task sequence. We present results that show the\nP-TNCN's ability to conduct zero-shot adaptation and online continual sequence\nmodeling.\n
This thesis aims to examine metapragmatic discourses on linguistic politeness illustrated in Korean language how-to literature. The primary task lies in contextualizing the native awareness of ene yeycel (linguistic politeness in Korean) within the interests or values of certain social groups. The first group, South Korean government-sanctioned agencies, led a linguistic campaign promoting a new standard speech model in 1992. Language professionals, the second group of social actors, produced popular language how-to literature, especially after the establishment of the hegemonic standard speech model. Both language standardizing policy and the participants in the how-to industry represent the cultural process of constructing language and social conventions. The “normative” culture of ene yeycel can be empowered and widely circulated, gaining wider social practice. Standardization of honorification came to the surface as a public issue along with a new “cultural policy” of the Ministry of Cultural Affairs in 1990. In this cultural-political circumstance, the social meaning of standardized honorification was rediscovered as indigenous culture, a group identity shared by Korean speakers. Positively valorizing honorification as linguistic and cultural tradition, the standardized model preserves the sophisticated use of honorifics and reinforces superior-inferior relationships. However, the standard model of ene yeycel can be subjective and arbitrary. Moreover, different styles are too easily proscribed as errors made by sloppy speakers. Language how-to literature produces more diversified interpretations than the standard speech manual. As language users are confronted with the challenges of finding the proper level of honorification, language how-to manuals provide justifications to help speakers prioritize linguistic norms when internalizing social relationships. Positive valorizations of honorification derive from a speaker's respect for the interlocutor's social status or personality. Negative valorizations of honorification view deferential politeness as a kind of discriminatory behaviour indexing power-difference. The positive or negative values of honorification are based on different concepts of ene yeycel and on different identifications of social relationships. Such conceptualizations rationalize whether speakers should support honorification or not, and lead them to discuss language use in current society.
Introduction:Iliac artery endofibrosis (IAE) is an uncommon disease, poorly studied pathology with devastating effects and different therapeutic approaches affecting young people who practise intensive sports, especially cyclists. The evolution of the process not only depends on the diagnosis and therapeutic action, but also on the acceptance and attitude of the patient and subsequent professional guidance. Case description:This is the case description of a professional triathlon athlete that had one previous iliac surgical revascularization for an IAE Iliac and was admitted in our department five times with subacute lower limb ischemia affecting both legs between 2013 and 2016. Clinical findings and image tests are reported, as well as medical procedures performed. Indications based on clinical, functional and imaging ratings were clear, but his professional activity was not completely abandoned. Finally, after four endovascular procedures with good immediate results, he was warned of the seriousness of the process since the etiopathogenic reason. At the present moment patient is asymptomatic, under routine controls, working as successful triathlon coach. Discussion and conclusion:The fact that an external mechanical stress is the reason of repeated iliac artery injury suggests that an open surgical approach correcting the external muscular compression or arterial deformation should be a definitive but also aggressive solution according to literature. However, endovascular procedures and new endovascular devices are an increasingly promising option with a very low surgical risk. No matter the revascularization performed, the persistence of sports intensive practice carries a high risk of recurrence. Sport practise cessation is mandatory in some cases in order to assure revascularization long-term patency, but also a well conducted professional orientation is needed to complete the therapeutic action.
Temporal models based on recurrent neural networks have proven to be quite powerful in a wide variety of applications, including language modeling and speech processing. However, to train these models, one relies on back-propagation through time, which entails unfolding the network over many time steps, making the process of conducting credit assignment considerably more challenging. Furthermore, the nature of back-propagation itself does not permit the use of non-differentiable activation functions and is inherently sequential, making parallelization of the underlying training process very difficult. In this work, we propose the Parallel Temporal Neural Coding Network, a biologically inspired model trained by the local learning algorithm known as Local Representation Alignment, that aims to resolve the difficulties and problems that plague recurrent networks trained by back-propagation through time. Most notably, this architecture requires neither unrolling nor the derivatives of its internal activation functions. We compare our model and learning procedure to other online back-propagation-through-time alternatives (which also tend to be computationally expensive), including real-time recurrent learning, echo state networks, and unbiased online recurrent optimization, and show that it outperforms them on sequence modeling benchmarks such as Bouncing MNIST, a new benchmark we call Bouncing NotMNIST, and Penn Treebank. Notably, our approach can, in some instances, even outperform full back-propagation through time itself as well as variants such as sparse attentive back-tracking. Furthermore, we present promising experimental results that demonstrate our model's ability to conduct zero-shot adaptation.
This paper explains transition dependency parsing approaches to build a dependency parser for Telugu language. Telugu treebank is given as an input to transition dependency parsers. One of the best transition dependency parser is the Malt parser. It is an independent system and it has nine methods to parse a sentence of any language. We have applied the treebank on all the methods of a Malt parser among which Arc-eager parser produces state-of-art results for Telugu language. Arc-eager method was produced LA (Label Accuracy) of 63%, UAS (Unlabeled Attachment Score) of 88.1% and LAS (Label Attachment Score) of 62.3%. In this paper we discuss a brief introduction of all Malt Parsing methods and an in detail explanation of Arc-eager dependency parsing.
Introduction. Nowadays the language of modern television broadcasts’ speechesis more and more in the focus of linguistic study. Special interest of given paper is comprisedby the variability in gender categorization of nouns as it occurs in speeches of Ukrainian TVprograms anchorpersons. The object of the paper is the choice of nouns in modern TV speech,that is distinguished due to the variability of grammatical category of the gender of nouns waysof realization.Purpose of the article is to analyze the nouns that have suffered changes in thegrammatical category of gender within the current language trends being implemented inthe broadcast of Ukrainian television. In addition, the aim is to outline some of the reasonsfor the emergence of such tendencies, the relative use of these nouns, the degree of codificationin modern lexicographic sources.Methods of research. The research is grounded on descriptive method, the method ofempiric analysis, immediate constituents’ analysis, and contextual analysis.Results. Studying the language used in contemporary Ukrainian TV speeches convincinglydemonstrated how formal-grammatical indicators of the category of gender of nouns in moderntelecommunication vary (shift), implemented in the following modifications: male genius femalegenus, female genus male genus. In addition, the review of codification in the dictionariesof the analyzed noun units shows that variational changes in the morphology of the noun andin particular, in its morphological and grammatical categories nowadays cause not onlythe dislocation of the current linguistic norm, but also tend to change the morphological norms. Conclusion. Analyzing the language used in speeches of informative and entertainingUkrainian TV programs’ anchorpersons, we have concluded that they prefer to choose differentgrammatical variations in favor of a specific counterpart to the grammatical category of the gender,often a revitalized or dialectal word used.
Introduction: The purpose of this investigation was to examine the independent and combined effects of caffeine (CAF) alone or as a part of a multi-ingredient pre-workout supplement (PWS) on resistance exercise performance in recreationally active males. Methods: In a single-blind, randomized, placebo (PLA) controlled, crossover design; 10 recreationally active males (20.5 ± 0.9; 178.9 ± 7.7 cm; 81.8 ± 11.5 kg) completed three laboratory visits, after determination of one repetition-maximum (1-RM) on the bench press and leg press, where they performed bench press and leg press to failure at a load of 70% 1-RM. Subjects were randomly assigned to ingest either one serving of a commercially available PWS (C4 Original, Cellucor, Bryan, TX, United States), a dosage-matched anhydrous CAF beverage (150 mg), or a taste- matched PLA beverage. Heart rate (HR), affect, rating of perceived exertion (RPE), and mood state was assessed 20 minutes pre and post-substance ingestion, and immediately after exercise.Results: Participants completed significantly more repetitions to failure (p = 0.006) and lifted significantly greater weight (p = 0.009) during Leg Press in the PWS and CAF conditions compared to the PLA condition. There was not a significant difference found between CAF and PWS trials (p > 0.05).Conclusions: This data suggests that both CAF and PWS may have a positive effect on exercise performance in leg press but was not effective in increasing muscular endurance in bench press. The commercially available PWS offered no additional ergogenic effects when compared to the CAF.
This paper carries out an empirical analysis of various dropout techniques for language modelling, such as Bernoulli dropout, Gaussian dropout, Curriculum Dropout, Variational Dropout and Concrete Dropout. Moreover, we propose an extension of variational dropout to concrete dropout and curriculum dropout with varying schedules. We find these extensions to perform well when compared to standard dropout approaches, particularly variational curriculum dropout with a linear schedule. Largest performance increases are made when applying dropout on the decoder layer. Lastly, we analyze where most of the errors occur at test time as a post-analysis step to determine if the well-known problem of compounding errors is apparent and to what end do the proposed methods mitigate this issue for each dataset. We report results on a 2-hidden layer LSTM, GRU and Highway network with embedding dropout, dropout on the gated hidden layers and the output projection layer for each model. We report our results on Penn-TreeBank and WikiText-2 word-level language modelling datasets, where the former reduces the long-tail distribution through preprocessing and one which preserves rare words in the training and test set.
Abstract In this article we provide a practical demonstration of how syntactically annotated corpora (treebanks), particularly the English Historical Parsed Corpora Series, can be used to investigate research questions with a diachronic depth and synchronic breadth that would not otherwise be possible. The phenomenon under investigation is split coordination, in which two parts of a conjoined constituent appear separated in the clause (e.g., and this is where my aunt lives and my uncle ). It affects every type of coordinated constituent (subject/object DPs, predicate and attributive ADJPs, ADVPs, PPs and DP objects of P) in Old English (OE); and it, or a superficially similar construction, occurs continuously throughout the attested period from approximately 800 to the present day. Despite its synchronic range and diachronic persistence, split coordination has received surprisingly little attention in the diachronic literature, with the exception of Perez Lorido’s (2009) limited study of split subjects in eight OE texts. Its modern counterpart is most frequently analysed as Bare Argument Ellipsis (BAE). Although the OE and Present-Day English constructions appear superficially similar, we show that not all of the OE data is amenable to a BAE analysis. We bring to bear different types of evidence (structural, discourse/performance effects, rate of change, etc.) to argue that split coordination in fact represents two different constructions, one of which remains stable over time while the other is lost in the post-Middle English period.
NLTK toolkit is an API platform built with Python language to interact with humans through natural language. The very first version of NLTK was released in 2005 (1.4.3), which was compatible with Python 2.4. The latest version was in September 2017 NLTK (3.2.5), which incorporated features like Arabic stemmers, NIST evaluation, MOSES tokenizer, Stanford segmenter, treebank detokenizer, verbnet, and vader, etc. NLTK was created in 2001 as a part of Computational Linguistic Department at the University of Pennsylvania. Since then it has been tested and developed. The important packages of this system are 1) corpus builder, 2) tokenizer, 3) collocation, 4) tagging, 5) parsing, 6) metrics, and 7) probability distribution system. Toolbox NLTK was built to meet four primary requirements: 1) Simplicity: An substantive framework for building blocks; 2) Consistency: Consistent interface; 3) Extensibility: Which can be easily scaled; and 4) Modularity: All modules are independent of each other.
espanolEn este articulo se lleva a cabo un analisis contrastivo de las cuestiones normativas, referentes a las categorias gramaticales de los determinantes y de los pronombres, incluidas en las gramaticas y en los diccionarios de la Real Academia Espanola. El corpus lo integran las diferentes ediciones de su obra gramatical (1771, 1796, 1854, 1870, 1883, 1911, 1917, el Esbozo de 1973 y la NGLE de 2009), asi como las veintitres ediciones de su obra lexicografica (desde 1780 hasta 2014). Se han examinado, clasificado y descrito los asuntos referentes a la norma, extraidos tras el cotejo de ambas obras a lo largo de la historia. Las principales conclusiones de este estudio revelan los siguientes datos: por una parte, son las gramaticas las que dedican una mayor atencion a los temas prescriptivos; por otra parte, en ciertas ocasiones, se observa una falta de coherencia entre las gramaticas y los diccionarios tanto en el seguimiento de la norma como en los criterios de correccion empleados; finalmente, con respecto a los diccionarios, sobresalen las ediciones manuales y la vigesima tercera edicion, dado su interes por recoger un mayor numero de alusiones al buen uso linguistico. EnglishIn this paper, a contrastive analysis of normative issues concerning the grammatical categories of determinants and pronouns included in the Spanish Royal Academy Grammars and Dictionaries is carried out. The corpus comprises the different editions of its grammatical work (1771, 1796, 1854, 1870, 1883, 1911, 1917, the 1973 Sketch and the NGLE of 2009), as well as the twenty-three editions of its lexicographical work (from 1780 to 2014). Issues related to the linguistic norm, which have been extracted from a comparison between the different editions throughout history, have been examined, classified and described. Results from this study reveal the following data: on the one hand, Grammars pay closer attention to prescriptive issues; on the other hand, a lack of coherence between Grammars and Dictionaries both in the follow-up of the linguistic norm and in the correction criteria used can be observed. In addition to that and with regard to dictionaries, the manual editions along with the 23rd edition are exceptional provided their interest in collecting a greater number of allusions to proper linguistic use.
Combination of customer needs and quantitative data is an idea to produce a various value. Therefore, this research proposed method for multidisciplinary design optimization of hub airport which optimized customer needs and quantitative data simultaneously. For customer needs, target of tourism and season were assigned into design variables and optimal weight was calculated by using SGD method. Design variables of quantitative data were number of transit, transportation fee and time from airport to World Heritage, and calculated value by using AHP which Monte Carlo simulation was applied for updating optimal value. To obtain optimal value for customer needs, review of tourism was extracted by text mining and count number of review in each World Heritage, word rating point was decided to calculate the weight of each World Heritage. Next, results of customer needs and quantitative data were normalized for collaborative optimization. Finally, the results showed a various optimal solution obtained for each design variable.
OBJECTIVE: To evaluate variables of tobacco health warnings associated with their emotional impact, the perception of smoking risks and the perceived effectiveness to avoid tobacco use. MATERIALS AND METHODS: Teenagers (151) and adults (168) evaluated 27 tobacco health warnings selected from the sets used on tobacco packages in Argentina and in other countries. A standardized affective rating-scale system and a structured questionnaire measured respectively the emotional impact (hedonic valence and emotional arousal), and the cognitive-behavioral attributions. The correlation between emotional and cognitive-behavioral evaluations was analyzed by age, sex, education level, smoker status,stage of quitting and susceptibility of non-smokers teenagers. RESULTS: Strong significant correlations between cognitivebehavioral and emotional assessments were observed. The warnings depicting graphic images of tobacco-related injuries and suffering were considered more valuable for tobacco. control, helping quitting and preventing initiation. CONCLUSIONS: Using graphic images with high emotional arousal is recommended for both adults and teenagers.
Detection and correction of errors and inconsistencies in "gold treebanks" are becoming more and more central topics of corpus annotation. The paper illustrates a new incremental method for enhancing treebanks, with particular emphasis on the extension of error patterns across different textual genres and registers. Impact and role of corrections have been assessed in a dependency parsing experiment carried out with four different parsers, whose results are promising. For both evaluation datasets, the performance of parsers increases, in terms of the standard LAS and UAS measures and of a more focused measure taking into account only relations involved in error patterns, and at the level of individual dependencies.
Recurrent neural nets (RNN) and convolutional neural nets (CNN) are widely used on NLP tasks to capture the long-term and local dependencies, respectively. Attention mechanisms have recently attracted enormous interest due to their highly parallelizable computation, significantly less training time, and flexibility in modeling dependencies. We propose a novel attention mechanism in which the attention between elements from input sequence(s) is directional and multi-dimensional (i.e., feature-wise). A light-weight neural net, "Directional Self-Attention Network (DiSAN)," is then proposed to learn sentence embedding, based solely on the proposed attention without any RNN/CNN structure. DiSAN is only composed of a directional self-attention with temporal order encoded, followed by a multi-dimensional attention that compresses the sequence into a vector representation. Despite its simple form, DiSAN outperforms complicated RNN models on both prediction quality and time efficiency. It achieves the best test accuracy among all sentence encoding methods and improves the most recent best result by 1.02% on the Stanford Natural Language Inference (SNLI) dataset, and shows state-of-the-art test accuracy on the Stanford Sentiment Treebank (SST), Multi-Genre natural language inference (MultiNLI), Sentences Involving Compositional Knowledge (SICK), Customer Review, MPQA, TREC question-type classification and Subjectivity (SUBJ) datasets.
We propose Efficient Neural Architecture Search (ENAS), a fast and inexpensive approach for automatic model design. In ENAS, a controller learns to discover neural network architectures by searching for an optimal subgraph within a large computational graph. The controller is trained with policy gradient to select a subgraph that maximizes the expected reward on the validation set. Meanwhile the model corresponding to the selected subgraph is trained to minimize a canonical cross entropy loss. Thanks to parameter sharing between child models, ENAS is fast: it delivers strong empirical performances using much fewer GPU-hours than all existing automatic model design approaches, and notably, 1000x less expensive than standard Neural Architecture Search. On the Penn Treebank dataset, ENAS discovers a novel architecture that achieves a test perplexity of 55.8, establishing a new state-of-the-art among all methods without post-training processing. On the CIFAR-10 dataset, ENAS designs novel architectures that achieve a test error of 2.89%, which is on par with NASNet (Zoph et al., 2018), whose test error is 2.65%.
This paper describes Stanford's system at the CoNLL 2018 UD Shared Task. We introduce a complete neural pipeline system that takes raw text as input, and performs all tasks required by the shared task, ranging from tokenization and sentence segmentation, to POS tagging and dependency parsing. Our single system submission achieved very competitive performance on big treebanks. Moreover, after fixing an unfortunate bug, our corrected system would have placed the 2 nd, 1 st, and 3 rd on the official evaluation metrics LAS, MLAS, and BLEX, and would have outperformed all submission systems on lowresource treebank categories on all metrics by a large margin. We further show the effectiveness of different model components through extensive ablation studies. * These authors contributed roughly equally.
When adding enhanced dependencies to an existing UD treebank, one can opt for heuristics that predict the enhanced dependencies on the basis of the UD annotation only. If the treebank is the result of conversion from an underlying treebank, an alternative is to produce the enhanced dependencies directly on the basis of this underlying annotation. Here we present a method for doing the latter for the Dutch UD treebanks. We compare our method with the UD -based approach of Schuster et al. (2018). While there are a number of systematic differences in the output of both methods, it appears these are the result of insufficient detail in the annotation guidelines and it is not the case that one approach is superior over the other in principle.
The performance of a machine translation system depends on the availability of bilingual lexical dictionary and completion of its word sense disambiguation performance. Word sense disambiguation plays a vital role in several applications such as machine translation, information retrieval and many other Natural Language Processing (NLP). In oder to construct a reliable machine translation system, not only consistent lexical database like WordNet (WN), but also word sense distinguished database, and bilingual lexical machine readable dictionary (MRD) are required to achieve an accurate translation. Since WN is for the English language, there have been several approcaches building WN like database for other languages, also known as multilingual WN, by linking and extending bilingual MRDs with synsets from WN. Lexical database like WN for Myanmar languae have been proposed in recent years which are bilingual lexical databases as the synsets from WN are merged with bilingual MRD. However, there are few limitations in constructing bilingual database based on MRD where thesauries are derived from WN yet there are no directed alignemnt of meanings between English and Myanmar language. In order to design a complete lexical database with correct parallel meanings for both languages, our proposed method composes a model which accurately matches the similar meanings between English words and Myanmar words by translating and comparing the word senses for both languages. The experimental result shows that the meanings for both languages are aligned as accurate as the manual alignment of the word senses for lexical database.
This paper describes our system (HIT-SCIR) submitted to the CoNLL 2018 shared task on Multilingual Parsing from Raw Text to Universal Dependencies. We base our submission on Stanford's winning system for the CoNLL 2017 shared task and make two effective extensions: 1) incorporating deep contextualized word embeddings into both the part of speech tagger and dependency parser; 2) ensembling parsers trained with different initialization. We also explore different ways of concatenating treebanks for further improvements. Experimental results on the development data show the effectiveness of our methods. In the final evaluation, our system was ranked first according to LAS (75.84%) and outperformed the other systems by a large margin.
We know very little about how neural language models (LM) use prior linguistic context.In this paper, we investigate the role of context in an LSTM LM, through ablation studies.Specifically, we analyze the increase in perplexity when prior context words are shuffled, replaced, or dropped.On two standard datasets, Penn Treebank and WikiText-2, we find that the model is capable of using about 200 tokens of context on average, but sharply distinguishes nearby context (recent 50 tokens) from the distant history.The model is highly sensitive to the order of words within the most recent sentence, but ignores word order in the long-range context (beyond 50 tokens), suggesting the distant past is modeled only as a rough semantic field or topic.We further find that the neural caching model (Grave et al., 2017b) especially helps the LSTM to copy words from within this distant context.Overall, our analysis not only provides a better understanding of how neural LMs use their context, but also sheds light on recent success from cache-based models.
We present the Uppsala system for the CoNLL 2018 Shared Task on universal dependency parsing. Our system is a pipeline consisting of three components: the first performs joint word and sentence segmentation; the second predicts part-ofspeech tags and morphological features; the third predicts dependency trees from words and tags. Instead of training a single parsing model for each treebank, we trained models with multiple treebanks for one language or closely related languages, greatly reducing the number of models. On the official test run, we ranked 7th of 27 teams for the LAS and MLAS metrics. Our system obtained the best scores overall for word segmentation, universal POS tagging, and morphological features.
The significance of tourism within the ASEAN region is recognised by multiple stakeholders. Presenting and promoting a distinctive image of tourism is a common agenda across the ten ASEAN nations. The aim of this study is to document and interpret the emotional connotations of ASEAN tourism slogans. Arguably, such messages provide an initial guide to the appeal and competitive advantages of each individual country. The study is underpinned by considering key ideas on destination positioning and the lexical analysis of emotions. By mining archival resources about word frequencies, synonyms and meanings, the positions of the slogans in an emotion space originally developed by Plutchik were compared and plotted. Joy, admiration and ecstasy were the dominant emotional connotations of most slogans. Thailand and Malaysia have the most distinctive tourism slogans, followed by Vietnam and Laos. Expressions used in the slogans for these four nations overlapped less with other countries across the families of emotion words.