Papers reviewed and determined not to be word norm studies. Use the flag icon to report errors or suggest re-inclusion.
16504 papers
Word ratings on affective dimensions are an important tool in psycholinguistic research. Traditionally, they are obtained by asking participants to rate words on each dimension, a time-consuming procedure. As such, there has been some interest in computationally generating norms, by extrapolating words’ affective ratings using their semantic similarity to words for which these values are already known. So far, most attempts have derived similarity from word co-occurrence in text corpora. In the current paper, we obtain similarity from word association data. We use these similarity ratings to predict the valence, arousal, and dominance of 14,000 Dutch words with the help of two extrapolation methods: Orientation towards Paradigm Words and k-Nearest Neighbors. The resulting estimates show very high correlations with human ratings when using Orientation towards Paradigm Words, and even higher correlations when using k-Nearest Neighbors. We discuss possible theoretical accounts of our results and compare our findings with previous attempts at computationally generating affective norms.
We propose a classification framework for semantic type identification of compounds in Sanskrit. We broadly classify the compounds into four different classes namely, Avyayībhāva, Tatpuruṣa, Bahuvrīhi and Dvandva. Our classification is based on the traditional classification system followed by the ancient grammar treatise Adṣṭādhyāyī, proposed by Pāṇini 25 centuries back. We construct an elaborate features space for our system by combining conditional rules from the grammar Adṣṭādhyāyī, semantic relations between the compound components from a lexical database Amarakoṣa and linguistic structures from the data using Adaptor Grammars. Our in-depth analysis of the feature space highlight inadequacy of Adṣṭādhyāyī, a generative grammar, in classifying the data samples. Our experimental results validate the effectiveness of using lexical databases as suggested by Amba Kulkarni and Anil Kumar, and put forward a new research direction by introducing linguistic patterns obtained from Adaptor grammars for effective identification of compound type. We utilise an ensemble based approach, specifically designed for handling skewed datasets and we %and Experimenting with various classification methods, we achieve an overall accuracy of 0.77 using random forest classifiers.
In accordance with the compositionality criterion and hierarchy principle of Rhetorical Structure Theory (RST), this study reframes each tree in the RST Discourse Treebank into three new dependency trees with ultimate nodes being clauses, sentences, and paragraphs, respectively, which also draw on an analogy between syntactic and discourse trees. Detailed percentages of various RST relations at the three granularity levels are examined, illuminating the discourse processes of organizing units of one granularity level into those of the next upper level and suggesting certain homogeneity and interaction across levels in the Treebank, particularly at the two upper levels. The study demonstrates the applicability of RST analysis between same-level terminal units. With unique analytical advantages, the newly constructed discourse dependency trees provide new research prospects.
We use reinforcement learning to learn tree-structured neural networks for computing representations of natural language sentences. In contrast with prior work on tree-structured models in which the trees are either provided as input or predicted using supervision from explicit treebank annotations, the tree structures in this work are optimized to improve performance on a downstream task. Experiments demonstrate the benefit of learning task-specific composition orders, outperforming both sequential encoders and recursive encoders based on treebank annotations. We analyze the induced trees and show that while they discover some linguistically intuitive structures (e.g., noun phrases, simple verb phrases), they are different than conventional English syntactic structures.
Estimates of the prevalence of sensitive attributes obtained through direct questions are prone to being distorted by untruthful responding. Indirect questioning procedures such as the Randomized Response Technique (RRT) aim to control for the influence of social desirability bias. However, even on RRT surveys, some participants may disobey the instructions in an attempt to conceal their true status. In the present study, we experimentally compared the validity of two competing indirect questioning techniques that presumably offer a solution to the problem of nonadherent respondents: the Stochastic Lie Detector and the Crosswise Model. For two sensitive attributes, both techniques met the “more is better” criterion. Their application resulted in higher, and thus presumably more valid, prevalence estimates than a direct question. Only the Crosswise Model, however, adequately estimated the known prevalence of a nonsensitive control attribute.
We present the results of the joint student response analysis (SRA) and 8th recognizing textual entailment challenge. The goal of this challenge was to bring together researchers from the educational natural language processing and computational semantics communities. The goal of the SRA task is to assess student responses to questions in the science domain, focusing on correctness and completeness of the response content. Nine teams took part in the challenge, submitting a total of 18 runs using methods and features adapted from previous research on automated short answer grading, recognizing textual entailment and semantic textual similarity. We provide an extended analysis of the results focusing on the impact of evaluation metrics, application scenarios and the methods and features used by the participants. We conclude that additional research is required to be able to leverage syntactic dependency features and external semantic resources for this task, possibly due to limited coverage of scientific domains in existing resources. However, each of three approaches to using features and models adjusted to application scenarios achieved better system performance, meriting further investigation by the research community.
The Universal Dependencies (UD) Project seeks to build a cross-lingual studies of treebanks, linguistic structures and parsing. Its goal is to create a set of multilingual harmonized treebanks that are designed according to a universal annotation scheme. In this paper, we report on the conversion of the Uyghur dependency treebank to a UD version of the treebank which we term the Uyghur Universal Dependency Treebank (UyDT). We present the mapping of the Uyghur dependency treebank’s labelling scheme to the UD scheme, along with a clear description of the structural changes required in this conversion.
We propose a framework to model human comprehension of discourse connectives. Following the Bayesian pragmatic paradigm, we advocate that discourse connectives are interpreted based on a simulation of the production process by the speaker, who, in turn, considers the ease of interpretation for the listener when choosing connectives. Evaluation against the sense annotation of the Penn Discourse Treebank confirms the superiority of the model over literal comprehension. A further experiment demonstrates that the proposed model also improves automatic discourse parsing.
Abstract Three studies examined gender differences in the effect of storytelling ability on perceptions of a person's attractiveness as a short‐term and long‐term romantic partner. In Study 1, information about a potential partner's storytelling ability was provided. Study 2 participants read a good or poor story supposedly written by a potential partner. Results suggested that only women's attractiveness assessments of men as a long‐term date increased for good storytellers. Storytelling ability did not affect men's ratings of women nor did it affect ratings of short‐term partners. Study 3 suggested that the effect of storytelling ability on long‐term attractiveness for male targets may be mediated by perceived status. Storytelling ability appears to increase perceived status and thus helps men attract long‐term partners.
Continuous spontaneous alternation behavior (SAB) in a Y-maze is used for evaluating working memory in rodents. Here, the design of an automated Y-maze equipped with three infrared optocouplers per arm, and commanded by a reduced instruction set computer (RISC) microcontroller is described. The software was devised for recording only true entries and exits to the arms. Experimental settings are programmed via a keyboard with three buttons and a display. The sequence of arm entries and the time spent in each arm and the neutral zone (NZ) are saved as a text file in a non-volatile memory for later transfer to a USB flash memory. Data files are analyzed with a program developed under LabVIEW® environment, and the results are exported to an Excel® spreadsheet file. Variables measured are: latency to exit the starting arm, sequence and number of arm entries, number of alternations, alternation percentage, and cumulative times spent in each arm and NZ. The automated Y-maze accurately detected the SAB decrease produced in rats by the muscarinic antagonist trihexyphenidyl, and its reversal by caffeine, having 100 % concordance with the alternation percentages calculated by two trained observers who independently watched videos of the same experiments. Although the values of time spent in the arms and NZ measured by the automated system had small discrepancies with those calculated by the observers, Bland-Altman analysis showed 95 % concordance in three pairs of comparisons, while in one it was 90 %, indicating that this system is a reliable and inexpensive alternative for the study of continuous SAB in rodents.
PURPOSE: The focus of this study was to examine the influence of fundamental frequency (F0) and vocal tract length (VTL) modifications on speaker gender recognition in cochlear implant (CI) recipients for different stimulus types. METHOD: Single words and sentences were manipulated using isolated or combined F0 and VTL cues. Using an 11-point rating scale, CI recipients and listeners with normal hearing rated the maleness/femaleness of the corresponding voice. RESULTS: Speaker gender ratings for combined F0 and VTL modifications were similar across all stimulus types in both CI recipients and listeners with normal hearing, although the CI recipients showed a somewhat larger ambiguity. In contrast to listeners with normal hearing, F0-VTL and F0-only modifications revealed similar ratings in the CI recipients when using words as stimuli. However, when sentences were used, a difference was found between F0-VTL-based and F0-based ratings. Modifying VTL cues alone did not affect ratings in the CI group. CONCLUSIONS: Whereas speaker gender ratings by listeners with normal hearing relied on combined VTL and F0 cues, CI recipients made only limited use of VTL cues, which might be one reason behind problems with identifying the speaker on the basis of voice. However, use of the voice cues depended on stimulus type, with the greater information in sentences allowing a more detailed analysis than single words in both listener groups.
Abstract We are investigating methods by which data from dependency syntax treebanks of ancient Greek can be applied to questions of authorship in ancient Greek historiography. From the Ancient Greek Dependency Treebank were constructed syntax words (sWords) by tracing the shortest path from each leaf node to the root for each sentence tree. This paper presents the results of a preliminary test of the usefulness of the sWord as a stylometric discriminator. The sWord data was subjected to clustering analysis. The resultant groupings were in accord with traditional classifications. The use of sWords also allows a more fine-grained heuristic exploration of difficult questions of text reuse. A comparison of relative frequencies of sWords in the directly transmitted Polybius book 1 and the excerpted books 9–10 indicate that the measurements of the two texts are generally very close, but when frequencies do vary, the differences are surprisingly large. These differences reveal that a certain syntactic simplification is a salient characteristic of Polybius’ excerptor, who leaves conspicuous syntactic indicators of his modifications.
The paper evaluates the differences between two currently leading annotation schemes for dependency treebanks. By relying on four treebanks, we demonstrate that the treatment of conjunctions and adpositions represents the core difference between the two schemes and that this impacts the topological properties of the linguistic networks induced from the treebanks. We also show that such properties are reflected in the performances of four probabilistic dependency parsers trained on the treebanks. L’articolo valuta le differenze tra i due principali schemi di annotazione a dipenden-ze in uso. Sulla base di quattro treebank, l’articolo dimostra che il trattamento delle congiunzioni e delle pre/postposizioni rappresenta la differenza principale tra i due schemi e che ciò comporta delle conseguenze sulle proprietà topologiche dei net-work indotti dalle treebank. Inoltre, si dimostra come tali proprietà siano riflesse nell’accuratezza di quattro parser probabilistici a dipendenze addestrati sulle treebank.
Semantic similarity has typically been measured across items of approximately similar sizes. As a result, similarity measures have largely ignored the fact that different types of linguistic item can potentially have similar or even identical meanings, and therefore are designed to compare only one type of linguistic item. Furthermore, nearly all current similarity benchmarks within NLP contain pairs of approximately the same size, such as word or sentence pairs, preventing the evaluation of methods that are capable of comparing different sized items. To address this, we introduce a new semantic evaluation called cross-level semantic similarity (CLSS), which measures the degree to which the meaning of a larger linguistic item, such as a paragraph, is captured by a smaller item, such as a sentence. Our pilot CLSS task was presented as part of SemEval-2014, which attracted 19 teams who submitted 38 systems. CLSS data contains a rich mixture of pairs, spanning from paragraphs to word senses to fully evaluate similarity measures that are capable of comparing items of any type. Furthermore, data sources were drawn from diverse corpora beyond just newswire, including domain-specific texts and social media. We describe the annotation process and its challenges, including a comparison with crowdsourcing, and identify the factors that make the dataset a rigorous assessment of a method’s quality. Furthermore, we examine in detail the systems participating in the SemEval task to identify the common factors associated with high performance and which aspects proved difficult to all systems. Our findings demonstrate that CLSS poses a significant challenge for similarity methods and provides clear directions for future work on universal similarity methods that can compare any pair of items.
Objectives: This study investigated the role of response style biases in the assessment of positive and negative affect in aging research; it addressed whether response styles (a) are associated with age-related changes in cognitive abilities, (b) lead to distorted conclusions about age differences in affect, and (c) reduce the convergent and predictive validity of affect measures in relation to health outcomes. Method: A multidimensional item response theory model was used to extract response styles from affect ratings provided by respondents to the psychosocial questionnaire (n = 6,295; aged 50-100 years) in the Health and Retirement Study (HRS). Results: The likelihood of extreme response styles (disproportionate use of "not at all" and "very much" response categories) increased significantly with age, and this effect was mediated by age-related decreases in HRS cognitive test scores. Removing response styles from affect measures did not alter age patterns in positive and negative affect; however, it consistently enhanced the convergent validity (relationships with concurrent depression and mental health problems) and predictive validity (prospective relationships with hospital visits, physical illness onset) of the affect measures. Discussion: The results support the importance of detecting and controlling response styles when studying self-reported affect in aging research.
We present a new, sizeable dataset of nounnoun compounds with their syntactic analysis (bracketing) and semantic relations. Derived from several established linguistic resources, such as the Penn Treebank, our dataset enables experimenting with new approaches towards a holistic analysis of noun-noun compounds, such as jointlearning of noun-noun compounds bracketing and interpretation, as well as integrating compound analysis with other tasks such as syntactic parsing.
Drawing on the work of Lawrence Abu Hamdan, a British-Lebanese artist and researcher currently based in Beirut, this essay examines the juridical and conceptual field of critical forensis which is situated at the juncture of security studies, art, and architecture. Abu Hamdan extends forensics to the area of “new audibilities,” with a focus on the politics of juridical hearing in situations of legal-identity profiling and voice authentication (the “shibboleth test”). Abu Hamdan's projects investigate how accent monitoring and audio surveillance, voice recognition, translation technologies, sovereign acts of listening, and court determinations of linguistic norms emerge as so many technical constraints on “freedom of speech,” itself a malleable term ascribed to discrepant claims and principles, yet taking on performative force in site-specific situations.
In this paper we study different types of Recurrent Neural Networks (RNN) for sequence labeling tasks. We propose two new variants of RNNs integrating improvements for sequence labeling, and we compare them to the more traditional Elman and Jordan RNNs. We compare all models, either traditional or new, on four distinct tasks of sequence labeling: two on Spoken Language Understanding (ATIS and MEDIA); and two of POS tagging for the French Treebank (FTB) and the Penn Treebank (PTB) corpora. The results show that our new variants of RNNs are always more effective than the others.
This paper presents the Universal Dependencies tagset (UD v1) as a new annotation scheme for Russian treebanks. The universal list of dependency relations was adopted and extended to comply with certain language-specific syntactic constructions. The tagset was validated, converting two Russian treebanks into the UD format, UD-Russian-SynTagRus and UD-Russian-Google.
The article presents methodological analyses of topical ideas of the famous modern linguist – E.Cosseriu. The authors argue that incorporation of theoretical ideas of E.Cosseriu could substentially extand the euristic potential of the conept of norm in the sphere of linguistics. The key to solvation of the problem lies in the necessity of changing of modern theoretical context of the question. The authors of the article consider the history of operationalization of the concept of norm in linguistics from the point of view of a specific hermeneutic approach in relation to other linguistic techniques. In the course of study of the role of linguistic norms in interaction of content and expression the article presents examples of extrapolation of the concept of norm from one discipline to another. In case of extrapolation of the concept of norm from the other disciplines into linguistic investigations the structural and functional dichotomy of language turns out that it is impossible to avoid considering spiritual as the world of objects, so that speech and language are considered as two different things. The rapid development of information technologies made possible to calculate many of the aspects of Humboldtian ideas. Computer statistics created conditions where the idea of "language picture of the world" and of the "inner form of the language" is gradually losing its original romantic charge and turns in a very trivial thing. Yet despite the fact that global standardization significantly enhances the processing and automatic addition of translation, which once gave beginning to hermeneutics the hermeneutic potential of linguistic norm still preseves many promissing prospects from the epistemological point of view. Key words: Linguistic norm and variability, E. Cosseriu, hermeneutics, euristic potential. В статье представлен методологический анализ актуальных идей известного современного лингвиста - E. Коссериу. Авторы утверждают, что введение теоретических идей E. Коссериу могло бы существенно расширить эвристический потенциал нормы в области лингвистики. Ключ к решению проблемы заключается в необходимости изменения современного теоретического контекста вопроса. Авторы статьи рассматривают историю ввода в действие понятия нормы в лингвистике с точки зрения конкретного герменевтического подхода по отношению к другим языковым методам. В ходе изучения роли языковых норм во взаимодействии содержания и выражения, в статье представлены примеры экстраполяции понятия нормы от одной дисциплины к другой. В случае экстраполяции понятия нормы с других дисциплин в лингвистические исследования структурно-функциональной дихотомии языка оказывается, что нельзя не рассматривать духовное как мир объектов, так как речь и язык рассматриваются как две разные вещи. Быстрое развитие информационных технологий сделало возможным для расчета многие аспекты гумбольдтовских идей. Статистика компьютера создала условия, в которых идея «языковой картины мира» и «внутренней формы языка» постепенно теряет свой первоначальный романтический заряд и превращается в очень тривиальную вещь. Тем не менее, несмотря на то, что глобальная стандартизация значительно улучшает обработку и автоматический перевод, который когда-то дал начало герменевтике, герменевтический потенциал языковой нормы хранит еще много перспектив с гносеологической точки зрения.Ключевые слова: лингвистическая норма и изменчивость, E.Коссериу, герменевтика, эвристический по- тенциал.
We investigate mutual benefits between syntax and semantic roles using neural network models, by studying a parsingSRL pipeline, a SRLparsing pipeline, and a simple joint model by embedding sharing. The integration of syntactic and semantic features gives promising results in a Chinese Semantic Treebank, demonstrating large potentials of neural models for joint parsing and semantic role labeling.
Recurrent Neural Network (RNN) is one of the most popular architectures used in Natural Language Processsing (NLP) tasks because its recurrent structure is very suitable to process variable-length text. RNN can utilize distributed representations of words by first converting the tokens comprising each text into vectors, which form a matrix. And this matrix includes two dimensions: the time-step dimension and the feature vector dimension. Then most existing models usually utilize one-dimensional (1D) max pooling operation or attention-based operation only on the time-step dimension to obtain a fixed-length vector. However, the features on the feature vector dimension are not mutually independent, and simply applying 1D pooling operation over the time-step dimension independently may destroy the structure of the feature representation. On the other hand, applying two-dimensional (2D) pooling operation over the two dimensions may sample more meaningful features for sequence modeling tasks. To integrate the features on both dimensions of the matrix, this paper explores applying 2D max pooling operation to obtain a fixed-length representation of the text. This paper also utilizes 2D convolution to sample more meaningful information of the matrix. Experiments are conducted on six text classification tasks, including sentiment analysis, question classification, subjectivity classification and newsgroup classification. Compared with the state-of-the-art models, the proposed models achieve excellent performance on 4 out of 6 tasks. Specifically, one of the proposed models achieves highest accuracy on Stanford Sentiment Treebank binary classification and fine-grained classification tasks.
This paper questions the nature of the communicative event that takes place in online contexts between doctors and web-users, showing computer-mediated linguistic norms and discussing the nature of the participants’ roles. Based on an analysis of 1005 posts occurring between doctors and the users of health service websites, I analyse how doctor–patient communication is affected by the medium and how health professionals overcome issues concerning the virtual medical visit. Results suggest that (a) online medical answers offer a different service from that expected by users, as doctors cannot always fulfill patient requests, and (b) net consultations use aspects of traditional doctor–patient exchange and yet present a language and a style that are affected by the computer-mediated environment. Additionally, it seems that this new form leads to a different model of doctor–patient relationship. The findings are intended to provide new insights into web-based discourse in doctor–patient communication and to demonstrate the emergence of a new style in medical communication.
Historical treebanks tend to be manually annotated, which is not surprising, since state-of-the-art parsers are not accurate enough to ensure high-quality annotation for historical texts. We test whether automatic parsing can be an efficient pre-annotation tool for Old East Slavic texts. We use the TOROT treebank from the PROIEL treebank family. We convert the PROIEL format to the CONLL format and use MaltParser to create syntactic pre-annotation. Using the most conservative evaluation method, which takes into account PROIEL-specific features, MaltParser by itself yields 0.845 unlabelled attachment score, 0.779 labelled attachment score and 0.741 secondary dependency accuracy (note, though, that the test set comes from a relatively simple genre and contains rather short sentences). Experiments with human annotators show that preparsing, if limited to sentences where no changes to word or sentence boundaries are required, increases their annotation rate. For experienced annotators, the speed gain varies from 5.80% to 16.57%, for inexperienced annotators from 14.61% to 32.17% (using conservative estimates). There are no strong reliable differences in the annotation accuracy, which means that there is no reason to suspect that using preparsing might lower the final annotation quality.
Lexical information, including surface word form and part-of-speech (POS) information, plays a crucial role when predicting ambiguous dependency relationships in dependency parsing. However, for resolving dependency ambiguities, surface word information may be too sparse, while POS information may be too coarse. Supertags, which are lexical templates that represent rich syntactic information, have been shown to provide effective features at an intermediate level on the coarse-to-fine scale. In this work, we present a supertag design framework that allows us to instantiate various supertag sets based on the dependency structures. Using this framework, we instantiate various supertag sets and utilize them as features in transition-based dependency parsing systems. Performing experiments on the Penn Treebank and Universal Dependencies data sets, we show that our supertags are effective for transition-based parsers in multilingual parsing as well as English parsing. The comparison of the results of the different supertag sets shows that it is crucial to incorporate the head directionality, head labels, and dependent possession information in supertags to improve the parser performance.
Anxiety disorders may not only be characterized by specific symptomatology (e.g., tachycardia) in response to the fearful stimulus (primary problem or first-level emotion) but also by the tendency to negatively evaluate oneself for having those symptoms (secondary problem or negative meta-emotion). An exploratory study was conducted driven by the hypothesis that reducing the secondary or meta-emotional problem would also diminish the fear response to the phobic stimulus. Thirty-three phobic participants were exposed to the phobic target before and after undergoing a psychotherapeutic intervention addressed to reduce the meta-emotional problem or a control condition. The electrocardiogram was continuously recorded to derive heart rate (HR) and heart rate variability (HRV) and affect ratings were obtained. Addressing the meta-emotional problem had the effect of reducing the physiological but not the subjective symptoms of anxiety after phobic exposure. Preliminary findings support the role of the meta-emotional problem in the maintenance of response to the fearful stimulus (primary problem).
Quantification is the machine learning task of estimating test-data class proportions that are not necessarily similar to those in training. Apart from its intrinsic value as an aggregate statistic, quantification output can also be used to optimize classifier probabilities, thereby increasing classification accuracy. We unify major quantification approaches under a constrained multi-variate regression framework, and use mathematical programming to estimate class proportions for different loss functions. With this modeling approach, we extend existing binary-only quantification approaches to multi-class settings as well. We empirically verify our unified framework by experimenting with several multi-class datasets including the Stanford Sentiment Treebank and CIFAR-10.
OBJECTIVES: Theoretical models of adult development suggest changes in emotion systems with age. This study determined how younger and older adults judged and classified 70 emotion terms that varied in valence and arousal, and that have been used in previous studies of adult aging and emotion. The terms were from the Positive and Negative Affect Schedule - Expanded (PANAS-X) and the (KS) affect scales. METHOD: Older (n = 32) and younger adults (n = 111) engaged in a card sort task which determined how the 70 emotion terms were classified (i.e. grouped) in relation to one another. Activation and valence ratings of emotion terms were collected. RESULTS: There were 17 age group differences in item ratings for activation and 19 for valence. Older adults tended to rate emotion terms and scales as more positive and activating than younger persons. Card sort data indicated similarity in conceptualizations of emotion terms across groups with exceptions for serene, sad, and lonely. CONCLUSIONS: Research that utilizes self-report emotion data from older and younger persons should consider how perceptions of emotion terms may vary systematically with age. The constructs of sadness, loneliness, and serene may be age-variant and necessitate age-based adjustments in assessment and intervention. Further, older adults may perceive some emotion terms to be more activating and positive than younger persons.
Normal-hearing listeners use acoustic cues in speech to interpret a speaker's emotional state. This study investigates the effect of hearing aids on the perception of the emotion dimensions arousal (aroused/calm) and valence (positive/negative attitude) in older adults with hearing loss. More specifically, we investigate whether wearing a hearing aid improves the correlation between affect ratings and affect-related acoustic parameters. To that end, affect ratings by 23 hearing-aid users were compared for aided and unaided listening. Moreover, these ratings were compared to the ratings by an age-matched group of 22 participants with age-normal hearing.For arousal, hearing-aid users rated utterances as generally more aroused in the aided than in the unaided condition. Intensity differences were the strongest indictor of degree of arousal. Among the hearing-aid users, those with poorer hearing used additional prosodic cues (i.e., tempo and pitch) for their arousal ratings, compared to those with relatively good hearing. For valence, pitch was the only acoustic cue that was associated with valence. Neither listening condition nor hearing loss severity (differences among the hearing-aid users) influenced affect ratings or the use of affect-related acoustic parameters. Compared to the normal-hearing reference group, ratings of hearing-aid users in the aided condition did not generally differ in both emotion dimensions. However, hearing-aid users were more sensitive to intensity differences in their arousal ratings than the normal-hearing participants.We conclude that the use of hearing aids is important for the rehabilitation of affect perception and particularly influences the interpretation of arousal.
This paper presents a novel latent variable recurrent neural network architecture for jointly modeling sequences of words and (possibly latent) discourse relations between adjacent sentences.A recurrent neural network generates individual words, thus reaping the benefits of discriminatively-trained vector representations.The discourse relations are represented with a latent variable, which can be predicted or marginalized, depending on the task.The resulting model can therefore employ a training objective that includes not only discourse relation classification, but also word prediction.As a result, it outperforms state-ofthe-art alternatives for two tasks: implicit discourse relation classification in the Penn Discourse Treebank, and dialog act classification in the Switchboard corpus.Furthermore, by marginalizing over latent discourse relations at test time, we obtain a discourse informed language model, which improves over a strong LSTM baseline.
This paper presents neural probabilistic parsing models which explore up to thirdorder graph-based parsing with maximum likelihood training criteria. Two neural network extensions are exploited for performance improvement. Firstly, a convolutional layer that absorbs the influences of all words in a sentence is used so that sentence-level information can be effectively captured. Secondly, a linear layer is added to integrate different order neural models and trained with perceptron method. The proposed parsers are evaluated on English and Chinese Penn Treebanks and obtain competitive accuracies.
We present the development and evaluation of a semantic analysis task that lies at the intersection of two very trendy lines of research in contemporary computational linguistics: (1) sentiment analysis, and (2) natural language processing of social media text. The task was part of SemEval, the International Workshop on Semantic Evaluation, a semantic evaluation forum previously known as SensEval. The task ran in 2013 and 2014, attracting the highest number of participating teams at SemEval in both years, and there is an ongoing edition in 2015. The task included the creation of a large contextual and message-level polarity corpus consisting of tweets, SMS messages, LiveJournal messages, and a special test set of sarcastic tweets. The evaluation attracted 44 teams in 2013 and 46 in 2014, who used a variety of approaches. The best teams were able to outperform several baselines by sizable margins with improvement across the 2 years the task has been run. We hope that the long-lasting role of this task and the accompanying datasets will be to serve as a test bed for comparing different approaches, thus facilitating research.
This paper describes FinnPos, an open-source morphological tagging and lemmatization toolkit for Finnish. The morphological tagging model is based on the averaged structured perceptron classifier. Given training data, new taggers are estimated in a computationally efficient manner using a combination of beam search and model cascade. The lemmatization is performed employing a combination of a rule-based morphological analyzer, OMorFi, and a data-driven lemmatization model. The toolkit is readily applicable for tagging and lemmatization of running text with models learned from the recently published Finnish Turku Dependency Treebank and FinnTreeBank. Empirical evaluation on these corpora shows that FinnPos performs favorably compared to reference systems in terms of tagging and lemmatization accuracy. In addition, we demonstrate that our system is highly competitive with regard to computational efficiency of learning new models and assigning analyses to novel sentences.
This paper is an extended description of SemEval-2014 Task 1, the task on the evaluation of Compositional Distributional Semantics Models on full sentences. Systems participating in the task were presented with pairs of sentences and were evaluated on their ability to predict human judgments on (1) semantic relatedness and (2) entailment. Training and testing data were subsets of the SICK (Sentences Involving Compositional Knowledge) data set. SICK was developed with the aim of providing a proper benchmark to evaluate compositional semantic systems, though task participation was open to systems based on any approach. Taking advantage of the SemEval experience, in this paper we analyze the SICK data set, in order to evaluate the extent to which it meets its design goal and to shed light on the linguistic phenomena that are still challenging for state-of-the-art computational semantic systems. Qualitative and quantitative error analyses show that many systems are quite sensitive to changes in the proportion of sentence pair types, and degrade in the presence of additional lexico-syntactic complexities which do not affect human judgements. More compositional systems seem to perform better when the task proportions are changed, but the effect needs further confirmation.
Most research on human fear conditioning and its generalization has focused on adults whereas only little is known about these processes in children. Direct comparisons between child and adult populations are needed to determine developmental risk markers of fear and anxiety. We compared 267 children and 285 adults in a differential fear conditioning paradigm and generalization test. Skin conductance responses (SCR) and ratings of valence and arousal were obtained to indicate fear learning. Both groups displayed robust and similar differential conditioning on subjective and physiological levels. However, children showed heightened fear generalization compared to adults as indexed by higher arousal ratings and SCR to the generalization stimuli. Results indicate overgeneralization of conditioned fear as a developmental correlate of fear learning. The developmental change from a shallow to a steeper generalization gradient is likely related to the maturation of brain structures that modulate efficient discrimination between danger and (ambiguous) safety cues.
Much previous research on multiword expressions (MWEs) has focused on the token- and type-level tasks of MWE identification and extraction, respectively. Such studies typically target known prevalent MWE types in a given language. This paper describes the first attempt to learn the MWE inventory of a “surprise” language for which we have no explicit prior knowledge of MWE patterns, certainly no annotated MWE data, and not even a parallel corpus. Our proposed model is trained on a treebank with MWE relations of a source language, and can be applied to the monolingual corpus of the surprise language to identify its MWE construction types.
Images: EEG montage used, beside international type; experiment overview; waveforms; difference topographies. For a more detailed view of the waveforms per group and electrode, visit: <b>https://pablobernabeu.shinyapps.io/export_files/</b><b>Abstract. </b>The engagement of sensory systems during word comprehension has been extensively documented; yet, the precise relevance of those remains unclear. We probed into this with an event-related potential (ERP) experiment which implemented the conceptual modality switch. This paradigm works as follows. In each trial, participants judge whether a property word can describe a concept word. However, the critical manipulation is the conceptual modality of the trials—e.g., haptic or visual—, as enabled by modality-normed stimuli. Switching across trials in different modalities, compared to maintaining a modality, incurs a switching cost. So far, experiments have measured this either on-line, from ERPs time-locked to the second word of the target trials, or off-line, from response times at the end of those trials. Problematically, both measurements fail to control a possible switch at the first word, as well as the semantic relation between the first and second words. In tackling the actual time frame of lexical and semantic access, we time-locked ERPs to the first word of target trials. Then, the experiment included different types of switch—from auditory to visual, and from haptic to visual—, which were compared to the non-switch—visual to visual. Further, we had a quick response group (<i>n</i> = 21), and a self-paced group (<i>n</i> = 21), alongside a few participants with no speed instructions (<i>n</i> = 5). The results, analyzed with mixed effects models, reveal ERP effects of modality-switching in four typical time windows between 160 and 750 ms after word onset. The effects are generally characterized by a more negative amplitude for modality-switching than not switching, and they arise with both types of switch, in both groups, and in anterior as well as posterior brain regions. In sum, the early start and broad scope of this effect suggest that perceptual simulation contributes fundamentally to word comprehension.
We compare different word embeddings from a standard window based skipgram model, a skipgram model trained using dependency context features and a novel skipgram variant that utilizes additional information from dependency graphs. We explore the effectiveness of the different types of word embeddings for word similarity and sentence classification tasks. We consider three common sentence classification tasks: question type classification on the TREC dataset, binary sentiment classification on Stanford's Sentiment Treebank and semantic relation classification on the SemEval 2010 dataset. For each task we use three different classification methods: a Support Vector Machine, a Convolutional Neural Network and a Long Short Term Memory Network. Our experiments show that dependency based embeddings outperform standard window based embeddings in most of the settings, while using dependency context embeddings as additional features improves performance in all tasks regardless of the classification method. Our embeddings and code are available at
The analysis of vocal expression is a critical endeavor for psychological and clinical sciences and is an increasingly popular application for computer–human interfaces. Despite this, and despite advances in the efficiency, affordability, and sophistication of vocal analytic technologies, there is considerable variability across studies regarding what aspects of vocal expression are studied. Vocal signals can be quantified in a myriad of ways, and their underlying structure, at least with respect to “macroscopic” measures from extended speech, is presently unclear. To address this issue, we evaluated the psychometric properties—notably, the structural and construct validity—of a systematically defined set of global vocal features. Our analytic strategy focused on (a) identifying redundant variables among this set, (b) employing principal components analysis (PCA) to identify nonoverlapping domains of vocal expression, (c) examining the degrees to which the vocal variables are modulated as a function of changes in speech task, and (d) evaluating the relationship between the vocal variables and cognitive (i.e., verbal fluency) and clinical (i.e., depression, anxiety, and hostility) variables. Spontaneous speech samples from 11 independent studies of young adults (>60 s in length), employing one of three different speaking tasks, were examined (N = 1,350). Confounding variables (i.e., sex, ethnicity) were statistically controlled for. The PCA identified six distinct domains of vocal expression. Collectively, vocal expression (defined in terms of these domains) was modulated as a function of speech task and was related to the cognitive and clinical variables. These findings provide empirically grounded implications for the study of vocal expression in psychological and clinical sciences.
Participant attentiveness is a concern for many researchers using Amazon’s Mechanical Turk (MTurk). Although studies comparing the attentiveness of participants on MTurk versus traditional subject pool samples have provided mixed support for this concern, attention check questions and other methods of ensuring participant attention have become prolific in MTurk studies. Because MTurk is a population that learns, we hypothesized that MTurkers would be more attentive to instructions than are traditional subject pool samples. In three online studies, participants from MTurk and collegiate populations participated in a task that included a measure of attentiveness to instructions (an instructional manipulation check: IMC). In all studies, MTurkers were more attentive to the instructions than were college students, even on novel IMCs (Studies 2 and 3), and MTurkers showed larger effects in response to a minute text manipulation. These results have implications for the sustainable use of MTurk samples for social science research and for the conclusions drawn from research with MTurk and college subject pool samples.
Stemming is a process of reducing a derivational or inflectional word to its root or stem by stripping all its affixes. It is been used in applications such as information retrieval, machine translation, and text summarization, as their pre-processing step to increase efficiency. Currently, there are a few stemming algorithms which have been developed for languages such as English, Arabic, Turkish, Malay and Amharic. Unfortunately, no algorithm has been used to stem text in Hausa, a Chadic language spoken in West Africa. To address this need, we propose stemming Hausa text using affix-stripping rules and reference lookup. We stemmed Hausa text, using 78 affix stripping rules applied in 4 steps and a reference look-up consisting of 1500 Hausa root words. The over-stemming index, under-stemming index, stemmer weight, word stemmed factor, correctly stemmed words factor and average words conflation factor were calculated to determine the effect of reference look-up on the strength and accuracy of the stemmer. It was observed that reference look-up aided in reducing both over-stemming and under-stemming errors, increased accuracy and has a tendency to reduce the strength of an affix stripping stemmer. The rationality behind the approach used is discussed and directions for future research are identified.
Images play an important role in the representation and acquisition of specialized knowledge. Not surprisingly, terminological knowledge bases (TKBs) often include images as a way to enhance the information in concept entries. However, the selection of these images should not be random, but rather based on specific guidelines that take into account the type and nature of the concept being described. This paper presents a proposal on how to combine the features of images with the conceptual propositions in EcoLexicon, a multilingual TKB on the environment. This proposal is based on the following: (1) the combinatory possibilities of concept types; (2) image types, such as photographs, drawings and flow charts; (3) morphological features or visual knowledge patterns (VKPs), such as labels, colours, arrows, and their effect on the functional nature of each image type. Currently, images are stored in association with concept entries according to the semantic content of their definitions, but they are not described or annotated according to the parameters that guided their selection, which would undoubtedly contribute to the systematization and automatization of the process. First, the images included in EcoLexicon were analyzed in terms of their adequateness, the semantic relations expressed, the concept types and their VKPs. Then, with these data, guidelines for image selection and annotation were created. The final aim is twofold: (1) to systematize the selection of images and (2) to start annotating old and new images so that the system can automatically allocate them in different concept entries based on shared conceptual propositions.
State-of-the-art prosody modelling in content-to-speech (CTS) applications still uses the same methodology to predict intonation cues as text-to-speech (TTS) applications, namely the analysis of the generated surface sentences with respect to part of speech, syntactic dependency relations and word order. On the other side, several theoretical studies argue that morphology, syntax, and information (or communicative) structure that organizes a given content (semantic or deep-syntactic structure) with respect to the intention of the speaker show a strong correlation with intonation. However, little empirical work based on sufficiently large corpora has been carried out so far to buttress this argumentation. We present empirical evidence for the Information Structure-Prosody correlation using the Wall Street Journal Penn Treebank corpus recorded by native American English speakers. Our experiments reach a prosody prediction accuracy of 80% using the hierarchical information structure from the Meaning-Text Theory, compared to 59% of the baseline.