Papers reviewed and determined not to be word norm studies. Use the flag icon to report errors or suggest re-inclusion.
18625 papers
In his later work in the philosophy of language Davidson analyses communicative exchanges and arrives at the startling and weighty conclusion that linguistic norms and conventions are entirely inessential to linguistic meaning. I argue that this inference is flawed: if we place the account of communication against the backdrop of Davidson's own views about radical interpretation, then it becomes evident that linguistic norms are an essential feature of the Davidsonian picture.
This paper presents preliminary investigations on the statistical parsing of French by bringing a complete evaluation on French data of the main probabilistic lexicalized and unlexicalized parsers first designed on the Penn Treebank. We adapted the parsers on the two existing treebanks of French (Abeillé et al., 2003; Schluter and van Genabith, 2007). To our knowledge, mostly all of the results reported here are state-of-the-art for the constituent parsing of French on every available treebank. Regarding the algorithms, the comparisons show that lexicalized parsing models are outperformed by the unlexicalized Berkeley parser. Regarding the treebanks, we observe that, depending on the parsing model, a tag set with specific features has direct influence over evaluation results. We show that the adapted lexicalized parsers do not share the same sensitivity towards the amount of lexical material used for training, thus questioning the relevance of using only one lexicalized model to study the usefulness of lexicalization for the parsing of French.
BACKGROUND: Time-limited group cognitive behavioral treatments (GCBT) for obsessive-compulsive disorder have demonstrated improvement in target symptoms. One small sample study of GCBT specifically for hoarding problems also showed benefit. This study examines the efficacy of a specialized GCBT for compulsive hoarding on a larger sample. METHODS: Thirty-two clients diagnosed with hoarding participated in five groups. Four groups met once weekly for 2 hour over 16 weeks (n=27) and one group met for 20 weeks (n=5). All participants had two individual 90-min home sessions. Self-report assessments were completed at baseline, mid-treatment, and post-treatment about hoarding behavior and related symptoms (e.g., depression). The sample was predominantly female, White, highly educated, unemployed, and not partnered/married; mean age was 53. A majority was diagnosed with major depressive disorder and obsessive-compulsive personality disorder. RESULTS: Participants showed significant improvement from pre- to post-treatment on the Saving Inventory Revised, Saving Cognitions Inventory, Clutter Image Rating, and Clinical Global Severity. The most recent group (n=8) that used a more formalized treatment and research protocol improved significantly more than did earlier members. CONCLUSION: This study demonstrates the feasibility and modest success of GCBT methods in improving hoarding symptoms. Group treatment may be especially valuable because of its cost-effectiveness, greater client access to trained clinicians, and reduction in social isolation and stigma linked to this problem. Further research is needed to improve the efficacy of GCBT methods for hoarding and to examine durability of change, predictors of outcomes, and processes that influence change.
There is strong evidence that human sentence processing is incremental, i.e., that structures are built word by word.Recent experiments show that the processor also predicts upcoming linguistic material on the basis of previous input.We present a computational model of human parsing that is based on a variant of tree-adjoining grammar and includes an explicit mechanism for generating and verifying predictions, while respecting incrementality and connectedness.An algorithm for deriving a lexicon from a treebank, a fully implemented parser, and a probability model for this formalism are also presented.We devise a linking function that explains processing difficulty as a combination of prefix probability (surprisal) and verification cost.The resulting model captures locality effects such as the subject/object relative clause asymmetry, as well as surprisal effects such as prediction in either...or constructions.
Jointly parsing two languages has been shown to improve accuracies on either or both sides. However, its search space is much bigger than the monolingual case, forcing existing approaches to employ complicated modeling and crude approximations. Here we propose a much simpler alternative, bilingually-constrained monolingual parsing, where a source-language parser learns to exploit reorderings as additional observation, but not bothering to build the target-side tree as well. We show specifically how to enhance a shift-reduce dependency parser with alignment features to resolve shift-reduce conflicts. Experiments on the bilingual portion of Chinese Treebank show that, with just 3 bilingual features, we can improve parsing accuracies by 0.6% (absolute) for both English and Chinese over a state-of-the-art baseline, with negligible (~6%) efficiency overhead, thus much faster than biparsing.
In this paper, we focus on facial displays, eye gaze and head tilts to express social dominance. In particular, we are interested in the interaction of different non-verbal cues. We present a study which systematically varies eye gaze and head tilts for five basic emotions and a neutral state using our own graphics and animation engine. The resulting images are then presented to a large number of subjects via a web-based interface who are asked to attribute dominance values to the character shown in the images. First, we analyze how dominance ratings are influenced by the conveyed emotional facial expression. Further, we investigate how gaze direction and head pose influence dominance perception depending on the displayed emotional state.
Treebanks are used for various purposes in language technology, but the wealth of data they contain can also be put to good use for the purpose of linguistic description and linguistic theory. To demonstrate this I will show how the treebank of the Spoken Dutch Corpus can be exploited to improve our understanding of what it is that distinguishes predicative complements from other types of complements. Section 1 shows why this distinction matters, section 2 provides a brief presentation of the treebank, section 3 gives a comprehensive survey of the intransitive predicate selecting verbs, based on the treebank data, section 4 presents a number of factors which can be used to differentiate the predicate selecting uses of the relevant verbs from their other uses, and section 5 summarizes the results. 1 Predicative complements Distinguishing a predicative complement from an object complement is easy in pairs like (1). (1) a. Fred is a plumber. b. Fred knows a plumber. The complement of the copula denotes a property which is attributed to the referent of the subject, and the copula itself is little more than a carrier of mood and tense. By contrast, the complement of know denotes an entity and the verb denotes a binary relation between that entity and the referent of the subject. Since the verb is the only element that overtly distinguishes (1a) from (1b), it might seem sufficient to draw the distinction, but the matter is more complex, since many verbs are used either way. The second complement of call and make, for instance, is predicative in (2), but not in (3). (2) a. Don’t call me a liar. b. They will make you chairman. 1This work is part of a larger project on the syntax and semantics of clauses with predicative complements. So far, it has yielded an HPSG style analysis of such clauses (Van Eynde 2008) and a semantic analysis of the copula, presented at the HPSG-2009 conference. Proceedings of the 19th Meeting of Computational Linguistics in the Netherlands Edited by: Barbara Plank, Erik Tjong Kim Sang and Tim Van de Cruys. Copyright c ©2009 by the individual authors.
The present study compared gender differences in directly reported and indirectly derived career preferences and tested the hypothesis that individuals' implicit preferences would show less gender-biased occupational choices than their directly elicited ones. Two hundred sixty-six visitors to a career-related Internet site were asked to (a) list 5 to 10 suitable occupations (the directly reported list) and (b) report their preferences in terms of 31 career-related aspects. The latter were used to produce a short list of promising occupational alternatives (the indirectly derived list), using the occupational database of an Internet-based career planning system. Each occupation in the database rated for sex dominance. The findings indicated that the sex dominance ratings of the occupations on the directly reported list accorded with the participants' gender for both men and women: Men's lists included mostly “masculine” occupations, whereas women's lists included mostly “feminine” occupations. This gender bias was significantly lower for the implicit lists. The difference between the directly reported and the indirectly derived lists was larger for women than for men, suggesting that the impact of stereotypes is more pronounced in women's than in men's directly reported career preferences.
Morphological segmentation breaks words into morphemes (the basic semantic units). It is a key component for natural language processing systems. Unsupervised morphological segmentation is attractive, because in every language there are virtually unlimited supplies of text, but very few labeled resources. However, most existing model-based systems for unsupervised morphological segmentation use directed generative models, making it difficult to leverage arbitrary overlapping features that are potentially helpful to learning. In this paper, we present the first log-linear model for unsupervised morphological segmentation. Our model uses overlapping features such as morphemes and their contexts, and incorporates exponential priors inspired by the minimum description length (MDL) principle. We present efficient algorithms for learning and inference by combining contrastive estimation with sampling. Our system, based on monolingual features only, outperforms a state-of-the-art system by a large margin, even when the latter uses bilingual information such as phrasal alignment and phonetic correspondence. On the Arabic Penn Treebank, our system reduces F1 error by 11% compared to Morfessor.
Dysfunctional emotional processing affects social functioning in patients with schizophrenia. However, the relationship between emotional perception and response in social interaction has not been elucidated. Twenty-seven patients with schizophrenia and 27 normal controls performed a virtual reality social encounter task in which they introduced themselves to avatars expressing happy, neutral, or angry emotions while verbal response duration and onset time were measured and perception of emotional valence and arousal, and state anxiety were rated afterwards. Self-reported trait-affective scale scores and the Positive and Negative Syndrome Scale (PANSS) ratings were also obtained. Patient group significantly underestimated the valence and arousal of angry emotions expressed by an avatar. While valence and arousal ratings of happy avatars were comparable between groups, patient group reported significantly higher state anxiety in response to happy avatars. State anxiety ratings significantly decreased from encounters with neutral to happy avatars in normal controls while no significant decrease was observed in the patient group. The Social Anhedonia Scale and PANSS negative symptom subscale scores (blunted affect, emotional withdrawal, and passive/ apathetic social withdrawal items) were significantly correlated with state anxiety ratings of the encounters with happy avatars. These results suggest that patients with schizophrenia have interference with the experience of pleasure in social interactions which may be associated with negative symptoms.
OBJECTIVES: This study is the first in a series designed to develop and norm new theoretically motivated sentence tests for children. The purpose was to examine the independent contributions of word frequency (i.e., how often words occur in language) and lexical density (the number of similar sounding words or "neighbors" to a target word) to the perception of key words in the new sentence set. DESIGN: Twenty-four children with normal hearing aged 5 to 12 yrs served as participants; they were divided into four equal age-matched groups. The stimuli consisted of 100 semantically neutral sentences that were 5 to 7 words in length. Each sentence contained 3 key words that were controlled for word frequency and lexical density. Words with few neighbors come from sparse neighborhoods, whereas words with many neighbors come from dense neighborhoods. The key words within a sentence belonged to one of the four lexical categories: (1) high-frequency sparse, (2) low-frequency dense, (3) high-frequency dense, and (4) low-frequency sparse. Participants were administered the sentence list and the 300 key words in isolation at 65 dB SPL. Each participant group was tested in spectrally matched noise at one of the four signal-to-noise ratios (SNRs -2, 0, 2, and 4 dB). The percent of words correctly identified was calculated as a function of SNR, key word context (sentences vs. words), and key word lexical category. RESULTS: SNR had a significant effect on the recognition of key words in sentences and in isolation; performance improved at higher SNRs. There were significant main effects of word frequency and lexical density as well as a significant interaction between the two lexical factors. In isolation, high-frequency words were recognized more accurately than low-frequency words. In both word and sentence contexts, sparse words yielded greater accuracy than dense words, irrespective of word frequency. There was a modest but significant negative correlation between lexical density and the recognition of words in isolation and in sentences. CONCLUSIONS: Word frequency and lexical density seem to influence word recognition independently in children with normal hearing. This is similar to earlier results in adults with normal hearing. In addition, there seems to be an interaction between the two factors, with lexical density being more heavily weighted than word frequency. These results give us further insight into the way children organize and access words from long-term lexical memory in a relational way. Our results showed that lexical effects were most evident at poorer SNRs. This may have important implications for assessing spoken-word recognition performance in children with sensory aids because they typically receive a degraded auditory signal.
German genitive attributes are usually tagged as such in treebanks. However, it is well known that this information is not sufficient for determining the type of relation between head nouns and attributes, as genitive attributes can express many different semantic relations. Various linguistic classifications have been worked out, but to my knowledge, nobody has so far proposed to apply this linguistic knowledge to a corpus. The challenge here is to come up with a classification that is both easy to verify and sufficiently fine-grained. Using earlier linguistic approaches as guidelines, I propose in this paper a detailed annotation scheme for German genitive attributes based on readily identifiable noun features. First insights from its application to the Smultron Treebank show that it is easy to distinguish between the proposed classes and that my classification of genitive attributes can be related to a more general semantic annotation level.
We present an implicit discourse relation classifier in the Penn Discourse Treebank (PDTB). Our classifier considers the context of the two arguments, word pair information, as well as the arguments' internal constituent and dependency parses. Our results on the PDTB yields a significant 14.1% improvement over the baseline. In our error analysis, we discuss four challenges in recognizing implicit relations in the PDTB.
Abstract We investigated how viewing positive and negative emotional stimuli influences the functional field of view. Two types of emotional pictures were used, representing either a negative emotion such as disgust or fear, or a positive emotion such vigour or excitement. The participants' task was to detect and identify a digit presented in any of the four corners of a picture on the display, while discriminating a letter in the centre of the display. We used two stimulus onset asynchronies (SOAs), of 500 and 3000 ms, between the picture and the digit. Performance was poorer in the negative condition than in the positive and non-emotional conditions. Performance in the 500 ms SOA condition was poorer than in the 3000 ms SOA condition. These results suggest that the functional field of view (FFOV) becomes narrower when people view negative emotional stimuli, whereas it does not change when viewing positive or neutral emotional stimuli. Keywords: Emotional stimuliThe functional field of viewSOA Acknowledgements This work was supported by a grant to the first author from the Research Fellowships of the Japan Society for the Promotion of Science for Young Scientists and Grant no. 15330156 from the Japan Society for the Promotion of the Science to the second author. The authors thank Sachio Nakamizo for his helpful comments on an earlier draft of this manuscript. Notes 1The IAPS slide numbers were as follows: positive, 4608, 4660, 5470, 8030, 8170, 8179, 8185, 8186, 8200, and 8490; negative, 1050, 1525, 2811, 3215, 3500, 3550, 6021, 6022, 6300, 6312, 6350, and 9921; neutral, 2102, 2393, 2396, 2579, 2850, 5530, 5731, 7034, 7179, 7205, 7490, and 7710. An ANOVA was performed on the pleasant ratings, arousal ratings, and the number of bytes of the compressed image file sizes with one factor of emotion (positive, negative, and neutral). In the pleasant ratings, the main effect of emotion was significant, F(2, 33) = 560.63, MSE=0.13, p<.001. A post hoc multiple comparison test using Ryan's method indicated that the positive picture (M=7.28) was higher than the neutral (M=5.25) or the negative (M=2.54) picture (p<.05), and the neutral picture was higher than the negative picture (p<.05). In the arousal ratings, the main effect of emotion was significant, F(2, 33) = 257.17, MSE=0.19, p<.001. A post hoc multiple-comparison test using Ryan's method indicated that the positive (M=6.64) or the negative picture (M=6.47) was higher than the neutral (M=3.04) picture (p<.05). In the number of bytes of the compressed image file sizes, a main effect of emotion was significant, F(2, 33) = 8.74, MSE=3641.56, p<.001. A post hoc multiple comparison test using Ryan's method indicated that the negative picture (M=112.83) was lower than the positive (M=187.00) or the neutral (M=211.33) pictures (p<.05).
This article describes a method for calculating the 'dependency distance' between the words in a text – i.e. the number of words that separate each word from the word on which it depends syntactically – and reports the results of applying this method to a Chinese treebank. This study shows that Chinese dependencies tend strongly to be governor-final and that the mean dependency distance of words is much higher for Chinese than for other languages that have been studied including English, German and Japanese. It is unclear whether this difference means that Chinese is syntactically more difficult to process.
Corpora present a basic informational source for varied NLP applications. Their construction becomes now days necessary. This paper presents a tool created to help syntactic tagging an Arabic Treebank. It is based on an already constructed grammar called ArabTAG which constitutes a representational method of Arabic grammatical rules using TAG formalism. This tagging tool is a node among a set of other pre-treatment steps as: morpho-syntactic analysis, grammatical tagging and sentence segmentation. It gets as input a sentence and helps to give it the appropriate syntactic tree in an incremental manner.
Dolgozatomban, amint arra a címből is lehet következtetni, az 1996 és 2005 között adatolható magyarországi börtönszlenget mutatom be, azt a csoportnyelvet, amelynek az átfogó tanulmányozása hazánkban eddig még nem történt meg. Munkámban büntetés-végrehajtási intézeteink fogvatartottjainak belső, informális nyelvhasználatának általános kérdéseivel foglalkozom, és a mai magyar börtönszleng szó- és kifejezéskészletének szótárba foglalásán túl kísérletet teszek a vizsgált csoportnyelv nyelvi-szociolingvisztikai leírására. \n \nCélkitűzésemet, a magyar börtönszleng átfogó tanulmányozását az indokolta, hogy a kutatás első éveiben olyan mennyiségű és minőségű, a nyelvtudományban eddig még nem tárgyalt adatokra bukkantam, amelyek érdemesnek mutatkoztak arra, hogy egy mélyebb, megtervezett szlengkutatás irányuljon erre a területre. \n \nElsősorban célom volt a magyar börtöszlenget feltárni, bemutatni keletkezését, funkcióját, működését, a szlenghasználó közösségben betöltött szerepét. Célom volt továbbá rámutatni nyelvi előzményeire, összevetni a már létező bűnözői nyelvi adatbázissal, vagyis a tolvajnyelv elemeivel, egyben definiálni helyét a magyar szlengkutatás területén. Kutatásom során mindvégig azt tartottam szem előtt, hogy hol van az ember a szlengben, így célom volt annak leírása is, mikor, milyen körülmények között motiváltak a vizsgált csoport tagjai szlenghasználatra. Ennek kiderítéséhez a nyelvi adatok feltárásán és rendszerezésén túl, a zárt közeg csoportjainak vizsgálatára is ki kellett terjeszteni a kutatást, megfigyelve a csoportszerveződés lehetőségeit és okait a börtöntársadalomban. \n \nAs already the title has suggested, my dissertation presents Hungarian prison slang as attested from 1996 to 2005. This group language has never seen an overall study in Hungary up to now. I deal with general questions of the internal and informal language use of the prisoners of penal institutions in my work and, together with rendering the words and expressions into a dictionary, I attempt at the linguistic and sociolinguistic description of the group language examined. \n \nMy objective, the overall study of Hungarian slang, is justified by the data being of such quantity and quality and having never been dealt with in linguistics that seemed worth to be examined by a deeper and planned slang research. \n \nMy main objective was to explore Hungarian prison slang, to present its origins, functions and operation as well as its role within slang user communities. My aim also was to show its linguistic predecessors, to compare it to an existing linguistic database of criminals, that is, to the elements of cant, and, together with it, to define its place in the area of Hungarian slang research. During my studies, I always kept man’s role in slang in mind so my objective was the description of the conditions among which the members of a researched group are motivated for slang usage. In order to learn about it, I had to extend my research to the examination of the groups of closed space, observing the possibilities and reasons within prison community.
OBJECTIVES: The present study in hypomanic and manic patients explored how amygdala responses to affective stimuli depend on the valence of the stimuli presented. METHODS: We compared 10 patients with 10 matched healthy control subjects. We measured blood oxygen level-dependent (BOLD) responses in the amygdala while subjects passively viewed photographs taken from the International Affective Picture System. After the fMRI session, subjects saw the pictures again and subjectively rated the emotional valence and intensity of each picture. RESULTS: Compared to healthy individuals, hypomanic or manic patients showed higher valence ratings in positive pictures and associated larger BOLD responses in the left amygdala during positive versus neutral picture viewing. This enhanced amygdala activation was correlated with Young Mania Rating Scale scores and with euphoric as opposed to irritable symptom presentation. CONCLUSIONS: Increased valence ratings and amygdala responses to positive affective stimuli may reflect a positive processing bias contributing to elevated mood states characteristic for euphoric mania.
Objective:To explore the characteristics of emotion cognitive processing and cognitive regulation in alexithymia.Methods:A total of 117 alexithymic subjects(TAS-20 scores ≥58)and 118 nonalexithymic subjects(TAS-20 scores ≤38)were selected with the Chinese version of 20-item Toronto Alexithymia Scale(TAS-20),and their scores on the Center for Epidemiologic Studies Depression Scale(CES-D)and Cognitive Emotion Regulation Questionnaire CERQ were compared.Then 51 alexithymic subjects and 54 nonalexithymic subjects were required to rate 120 affective pictures to three dimensions(valence,arousal and dominant).Results:(1)Compared with nonalexithymic group,alexithymic group got higher scores in negative coping dimension[(47.3±5.9) vs.(41.9±5.9),P0.001],while got lower scores in positive coping [(65.2±7.7) vs.(71.1±7.3),P0.001].(2)In valence rating,alexithymic group gave lower score for positive pictures[(7.0±1.0) vs.(7.7±1.0),P0.001]and higher score for negative pictures [(2.4±1.0) vs.(1.4±1.0),P0.001]than nonalexithymic group.In arousal dimension,alexithymic group gave lower scores to both positive pictures and negative pictures than nonalexithymic group[(6.3±1.2) vs.(6.8±1.1),(6.4±1.5) vs.(7.2±1.4);P0.01].Conclusion:Alexithymia subjects have deficits in emotion cognitive processing and emotion cognitive regulation.
The present paper outlines an ongoing project of annotation of the extended nominal coreference and the bridging anaphora in the Prague Dependency Treebank. We describe the annotation scheme with respect to the linguistic classification of coreferential and bridging relations and focus also on details of the annotation process from the technical point of view. We present methods of helping the annotators -- by a pre-annotation and by several useful features implemented in the annotation tool. Our method of the inter-annotator agreement is focused on the improvement of the annotation guidelines; we present results of three subsequent measurements of the agreement.
The lexical development system OntoNet is introduced, which includes a browser and an editor for the WordNet 3.0 database. The aim of the OntoNet project is to provide a comfortable and up-to-date access to the lexical database for the modification of WordNet or the development of new wordnets.
We describe for dependency parsing an annotation adaptation strategy, which can automatically transfer the knowledge from a source corpus with a different annotation standard to the desired target parser, with the supervision by a target corpus annotated in the desired standard. Furthermore, instead of a hand-annotated one, a projected treebank derived from a bilingual corpus is used as the source corpus. This benefits the resource-scarce languages which haven't different handannotated treebanks. Experiments show that the target parser gains significant improvement over the baseline parser trained on the target corpus only, when the target corpus is smaller.
The aim of the project entitled "Computerized Historical Linguistic Database of Latin Inscriptions of the Imperial Age" (http://lldb.elte.hu/) is to develop and digitally publish a fundamental computerized historical linguistic database that incorporates and treats the Vulgar Latin material of the Latin inscriptions from a specific group of the European provinces of the Roman Empire in the first phase. This will, on the one hand, allow for a more thorough study of the regional changes and the diversity of the Latin language of the Imperial Age. On the other hand, it could also serve as a basis for subsequent international co-operation, in the course of which further work on the computerized historical linguistic database may be executed. This paper intends to present the past and the present, as well as the future possibilities of this Database.
We describe the Hindi Discourse Relation Bank project, aimed at developing a large corpus annotated with discourse relations. We adopt the lexically grounded approach of the Penn Discourse Treebank, and describe our classification of Hindi discourse connectives, our modifications to the sense classification of discourse relations, and some crosslinguistic comparisons based on some initial annotations carried out so far.
Standard English PoS-taggers generally involve tag-assignment (via dictionary-lookup etc) followed by tag-disambiguation (via a context model, e.g. PoS-ngrams or Brill transformations). We want to PoS-tag our Arabic Corpus, but evaluation of existing PoS-taggers has highlighted shortcomings; in particular, about a quarter of all word tokens are not assigned a fully correct morphological analysis. Tag-assignment is significantly more complex for Arabic. An Arabic lemmatiser program can extract the stem or root, but this is not enough for full PoS-tagging; words should be decomposed into five parts: proclitics, prefixes, stem or root, suffixes and postclitics. The morphological analyser should then add the appropriate linguistic information to each of these parts of the word; in effect, instead of a tag for a word, we need a subtag for each part (and possibly multiple subtags if there are multiple proclitics, prefixes, suffixes and postclitics). Many challenges face the implementation of Arabic morphology, the rich “root-and-pattern” nonconcatenative (or nonlinear) morphology and the highly complex word formation process of root and patterns, especially if one or two long vowels are part of the root letters. Moreover, the orthographic issues of Arabic such as short vowels ( ), Hamzah (ء أ إ ؤ ئ), Taa’ Marboutah ( ة ) and Ha’ ( ه ), Ya’ ( ي ) and Alif Maksorah( ى ), Shaddah ( ) or gemination, and Maddah ( آ ) or extension which is a compound letter of Hamzah and Alif ( أا ). Our morphological analyzer uses linguistic knowledge of the language as well as corpora to verify the linguistic information. To understand the problem, we started by analyzing fifteen established Arabic language dictionaries, to build a broad-coverage lexicon which contains not only roots and single words but also multi-word expressions, idioms, collocations requiring special part-of-speech assignment, and words with special part-of-speech tags. The next stage of research was a detailed analysis and classification of Arabic language roots to address the “tail” of hard cases for existing morphological analyzers, and analysis of the roots, word-root combinations and the coverage of each root category of the Qur’an and the word-root information stored in our lexicon. From authoritative Arabic grammar books, we extracted and generated comprehensive lists of affixes, clitics and patterns. These lists were then cross-checked by analyzing words of three corpora: the Qur’an, the Corpus of Contemporary Arabic and Penn Arabic Treebank (as well as our Lexicon, considered as a fourth cross-check corpus). We also developed a novel algorithm that generates the correct pattern of the words, which deals with the orthographic issues of the Arabic language and other word derivation issues, such as the elimination or substitution of root letters.
Most treebank work in the past has focused on European and Asian languages. The Wikipedia Treebank page lists treebanks (or treebank projects) for about 20 modern European languages (ranging from Basque to Swedish), five Asian languages (Chinese, Japanese, Hindi, Korean, Thai), two ancient languages (Greek and Latin), plus Arabic and Hebrew. Almost no treebanking work has been done on African or American indigenous languages.1 In the past we have explored parallel treebanks for English, German and Swedish [7]. Now we would like to explore to what extent our tools and guidelines will work when we include a very different language, Quechua, for which only few NLP resources exist. Since Quechua is spoken in Latin America, Spanish as parallel language is a natural choice. We have first compiled a parallel corpus Quechua Spanish. We have then stepwise analyzed and annotated the Quechua and the Spanish texts. For Spanish we have used the treebanking guidelines developed by [8]. As for Quechua there were no such guidelines so that we had to experiment with finding the appropriate grammar formalism and develop our own guidelines. In this paper we describe the characteristics of Quechua and our steps towards its morphological and syntactic annotation. We argue for Role and Reference Grammar as a suitable grammar formalism. We briefly describe how we annotated the parallel Spanish texts and demonstrate how we plan to align the Quechua with the Spanish trees.
The first edition of a speech corpus with a speech reconstruction layer (edited transcript). The project of speech reconstruction of Czech and English has been started at UFAL together with the PIRE project in 2005, and has gradually grown from ideas to (first) annotation specification, annotation software and actual annotation. It is part of the Prague Dependency Treebank family of annotated corpus resources and tools, to which it adds the spoken language layer(s).
The NLP community has shown a renewed interest in deeper semantic analyses, among them automatic recognition of semantic relations in text. We present the development and evaluation of a semantic analysis task: automatic recognition of relations between pairs of nominals in a sentence. The task was part of SemEval-2007, the fourth edition of the semantic evaluation event previously known as SensEval. Apart from the observations we have made, the long-lasting effect of this task may be a framework for comparing approaches to the task. We introduce the problem of recognizing relations between nominals, and in particular the process of drafting and refining the definitions of the semantic relations. We show how we created the training and test data, list and briefly describe the 15 participating systems, discuss the results, and conclude with the lessons learned in the course of this exercise.
Annotated data have recently become more important, and thus more abundant, in computational linguistics. They are used as training material for machine learning systems for a wide variety of applications from Parsing to Machine Translation (Quirk et al., 2005). Dependency representation is preferred for many languages because linguistic and semantic information is easier to retrieve from the more direct dependency representation. Dependencies are relations that are defined on words or smaller units where the sentences are divided into its elements called heads and their arguments, e.g. verbs and objects. Dependency parsing aims to predict these dependency relations between lexical units to retrieve information, mostly in the form of semantic interpretation or syntactic structure. Parsing is usually considered as the first step of Natural Language Processing (NLP). To train statistical parsers, a sample of data annotated with necessary information is required. There are different views on how informative or functional representation of natural language sentences should be. There are different constraints on the design process such as: 1) how intuitive (natural) it is, 2) how easy to extract information from it is, and 3) how appropriately and unambiguously it represents the phenomena that occur in natural languages. In this article, a review of statistical dependency parsing for different languages will be made and current challenges of designing dependency treebanks and dependency parsing will be discussed.
The Postmodern culture today breaks down historically a solid barrier, produced in the modern age, between the language and its users, signs and realities and subject and his object. This phenomenon brings about some translation problems at the same time; interpretative diversity, linguistic derivation in the mass media, permanent reproduction of translated text and meaning`s continuity, definition of translator etc. I examine these problems through a theoretical approach to the mediative nature of the act of translation. I stress on two inevitable aspects of the postmodern linguistic tendencies: connotation generalized in our translation activities and its re-mediative culture. Focusing on a cycling perspective of our translating activities, I could reach the conclusion that the translated text have no relation with the denotative meanings which have been comprised closed and original with the realities from the modern age. A corrected model could be proposed. I call this cycle model of translating process which comprises ① Mediation or Creation ② Re-mediation or Translation ③ Re-re-mediation or Comprehension ④ Verification, Deduction or Correction. Each activity has not only its own translating process but its circulated role for the time in which everyone can participate equally as a reader, a sender, a translator and an individual who makes its contextual needs and desires. This model could show that the translating process is not for fixing a linguistic sign to some closed meanings but for expanding its pertinent meanings to diverse situations. Translating activity is not for making a linguistic norm by itself, but for making appropriate communication with as much of the population as possible.
This research concerns linguistic variation and Portuguese teaching at school. It is assumed that linguistic pattern is conceived by teachers as an homogeneous norm, so that it is incompatible with linguistic norms students face with in usual text reading and writing activities. Taking into consideration official evaluation of didactic books, scholar reading activities, and sociolinguistic results that prove school interference at students writing performance, this article proposes that Portuguese classes should present variation as complex continua in which is displayed a plurality of norms.
Background: One of the most debated issues in the cognitive neuroscience of language is whether distinct semantic domains are differentially represented in the brain. Clinical studies described several anomic dissociations with no clear neuroanatomical correlate. Neuroimaging studies have shown that memory retrieval is more demanding for proper than common nouns in that the former are purely arbitrary referential expressions. In this study a semantic relatedness paradigm was devised to investigate neural processing of proper and common nouns. Methodology/Principal Findings: 780 words (arranged in pairs of Italian nouns/adjectives and the first/last names of well known persons) were presented. Half pairs were semantically related (''Woody Allen'' or ''social security''), while the others were not (''Sigmund Parodi'' or ''judicial cream''). All items were balanced for length, frequency, familiarity and semantic relatedness. Participants were to decide about the semantic relatedness of t)
Un assunto sempre più condiviso nell’ambito degli studi sull’acquisizione sia di L1 che di L2 è che l’evidenza empirica privilegiata debba essere rappresentata da corpora di produzioni scritte o orali degli apprendenti, estensivamente annotate a molteplici livelli di rappresentazione linguistica. Più in generale, corpora lemmatizzati e annotati a livello morfosintattico fanno ormai parte dello strumentario comune del linguista. Accanto ad essi, si fa però strada l’esigenza di disporre di risorse testuali più sofisticate dal punto di vista delle modalità di esplorazione linguistica, come ad esempio corpora annotati a livello sintattico (le cosiddette treebank). Questi consentono infatti di osservare i processi di convergenza degli apprendenti verso la lingua “obiettivo” anche a livello di specifici tratti grammaticali astratti o di macro-strutture linguistiche.
The magnetic resonance imaging (MRI) machine itself has an impact on the likelihood of obtaining successful measurements of brain size in certain groups of subjects. The differential selection and attrition in both cross-sectional and longitudinal designs, therefore, indicate that the MRI coincidentally serves as a screen for the anatomical structure of the brains that are successfully scanned. This screening effect introduces confounds in experiments whose very hypotheses are focused on comparing anatomical differences in subjects who differ, for example, in their reactions to anxiety-inducing situations. Here, behavioral interventions and possible statistical models are presented in order to reduce attrition and other effects of the confounds introduced by the MRI measurement process in research. Child and adolescent research—particularly in the attention-deficit/hyperactivity disorder research area—is used as an example to clarify and delineate the general research principles presented in the present article.
This dissertation examines the relationship between gender and language in Japanese through the often ignored lens of sexuality. Although linguists are increasingly examining these issues for American gay, lesbian, and bisexual speakers, little similar research has been done in Japan. Lesbians, in particular, are relatively invisible in Japanese society. Examining these women, who do not fit neatly into the hegemonic gender ideology, illuminates how speakers can project a specific identity by displaying or rejecting prescriptive gender-specific linguistic norms of Japanese.I analyzed data recorded from interviews with both Japanese lesbian/bisexual and heterosexual women, looking for differences in frequency and range of use of pronouns and sentence-final particles and for phonetic differences in terms of average pitch height and width. I also considered the results of a perception experiment undertaken to investigate the effect of pitch height and width on Japanese speakers' perceptions of sexuality.Although Japanese speakers were generally unable to identify a cohesive lesbian stereotype, especially in terms of language use, the perception experiment indicated that both average pitch height and width significantly affect judgments on whether a voice sounds lesbian or heterosexual. Tokens judged to be lesbian were also judged to be more masculine and less emotional than those judged to be heterosexual. Analysis of the interview data showed that lesbian participants produced an average pitch height that was significantly lower than that of heterosexual participants. In terms of gendered morphemes, lesbians were significantly more likely to use masculine morphemes than heterosexual women, both for sentence-final particles and first-person pronouns, and were significantly less likely to use the feminine first-person pronoun atashi. Finally, correlations showed that speakers who instantiate gender through the use of gendered-morphemes also do so through manipulations of pitch.Although Japanese lesbians are still fairly closeted and interviewees maintained that there are no cultural stereotypes for this group, significant differences in pitch and gendered-morpheme usage were still apparent. These lesbian/bisexual women did not appear to be mimicking men's language, but instead seemed to be rejecting hegemonic femininity and many of the cultural and linguistic stereotypes that accompany it.
Computer-aided Acquisition of Semantic Knowledge (CASK) is aimed at describing a number of semantic fields of a few European languages using data mining techniques elaborated within the framework of the new paradigm of computation known as Knowledge Discovery in Databases (KDD). CASK's motivation is to dig deeper in order to find building blocks which could be used in various sophisticated ways. The project is interdisciplinary involving scientific cooperation of experts in linguistics with information engineers. The task of linguists consists in an interactive (computer-aided) discovery of ontology-based definitions of feature structures using the SEMANA (Semantic Analyser) software which was designed especially in order to build linguistic databases with semantic knowledge.
English is spoken worldwide by both native (L1) and nonnative (L2) speakers. It is therefore imperative to establish how easily L1 and L2 speakers understand each other. We know that L1 listeners adapt to foreign-accented speech very rapidly (Clarke & Garrett, 2004), and L2 listeners find L2 speakers (from matched and mismatched L1 backgrounds) as intelligible as native speakers (Bent & Bradlow, 2003). But foreign-accented speech can deviate widely from L1 pronunciation norms, for example when adult L2 learners experience difficulties in producing L2 phonemes that are not part of their native repertoire (Strange, 1995). For instance, Italian L2 learners of English often lengthen the lax English vowel /I/, making it sound more like the tense vowel /i/ (Flege et al., 1999). This blurs the distinction between words such as bin and bean. Unless listeners are able to adapt to this kind of pronunciation variance, it would hinder word recognition by both L1 and L2 listeners (e.g., /bin/ could mean either bin or bean). In this study we investigate whether Italian-accented English interferes with on-line word recognition for native English listeners and for nonnative English listeners, both those where the L1 matches the speaker accent (i.e., Italian listeners) and those with an L1 mismatch (i.e., Dutch listeners). Second, we test whether there is perceptual adaptation to the Italian-accented speech during the experiment in each of the three listener groups. Participants in all groups took part in the same cross-modal priming experiment. They heard spoken primes and made lexical decisions to printed targets, presented at the acoustic offset of the prime. The primes, spoken by a native Italian, consisted of 80 English words, half with /I/ in their standard pronunciation but mispronounced with an /i/ (e.g., trick spoken as treek), and half with /i/ in their standard pronunciation and pronounced correctly (e.g., treat). These words also appeared as targets, following either a related prime (which was either identical, e.g., treat-treat, or mispronounced, e.g., treek-trick) or an unrelated prime. All three listener groups showed identity priming (i.e., faster decisions to treat after hearing treat than after an unrelated prime), both overall and in each of the two halves of the experiment. In addition, the Italian listeners showed mispronunciation priming (i.e., faster decisions to trick after hearing treek than after an unrelated prime) in both halves of the experiment, while the English and Dutch listeners showed mispronunciation priming only in the second half of the experiment. These results suggest that Italian listeners, prior to the experiment, have learned to deal with Italian-accented English, and that English and Dutch listeners, during the experiment, can rapidly adapt to Italian-accented English. For listeners already familiar with a particular accent (e.g., through their own pronunciation), it appears that they have already learned how to interpret words with mispronounced vowels. Listeners who are less familiar with a foreign accent can quickly adapt to the way a particular speaker with that accent talks, even if that speaker is not talking in the listeners’ native language.
This paper proposes the syntactic category prediction for improving translation quality. In parsing using sentence segmentation, the segments are separately parsed and then the parsing results of each segment are combined to generate a global sentence structure. The syntactic category prediction guides the parser to identify relationships among segments and to select the correct parsing results for each segment. We design features for predicting syntactic categories and generate decision trees for the prediction using training data from the Penn Treebank. In experiment, we show the prediction accuracy and comparison results with the prediction by human-built rules, heuristic probability function, and neural networks. Also, we present how much the category prediction contributes to improving translation quality.
Parsing is an important process of Natural Language Processing (NLP) and Computational Linguistics which is used to understand the syntax and semantics of natural language sentences confined to the grammar. Parsing models need syntax and semantic coverage for better interpretation of natural language sentences. Though statistical parsing with trigram language models gives better performance through tri-gram probabilities and large vocabulary size, it has some disadvantages like lack of support in syntax, free ordering of words and long distance relationship which are the challenging features of the Tamil language. Grammar based structural parsing provides solutions to some extent. To overcome these disadvantages, structural component is to be involved in statistical approach which results in hybrid models like phrase and dependency models. To add the structural component, balance the vocabulary size and meet the challenging features, lexicalized and statistical parsing (LSP) is to be employed with the assistance of hybrid models. To incorporate all the features in complex and large sentences, phrase structure model may not be suitable to a larger extent. When dependency relations are applied among words, direct relationships can be established. Lexicalized and statistical parsing of natural language text in Tamil language using dependency model will give better performance than using phrase structure model. New part of speech (POS) and dependency tag sets for Tamil language have been Treebank has been developed with 326 sentences which comprises more than 5000 words with manual annotation. It has been extended to 1000 sentences using bootstrapping and manual correction and used to train the dependency model. This LSP with dependency model provides better results and covers all the features of Tamil language.
Alternating Structure Optimization (ASO) is a recently proposed linear Multitask Learning algorithm. Although its effective has been verified in both semi-supervised as well as supervised methods, yet they necessitate taking external resource as a prerequisite. Therefore, feasibility of employing ASO to further improve the performance merely rests on the labeled data on hand proves to be a task deserving close scrutiny. Catering to this challenging while untapped problem, this paper presents a novel application of ASO to the subtask of Shallow Semantic Parsing: Chunking. Our experiments on Chinese Treebank 5.0 present promising result in chunk analysis, and the error rate is reduced by 5.72%, proposing a profound way to further improve the performance.
This paper studies the nature of the BEI-construction in Cantonese, with Mandarin as the standard language of comparison. Although the BEI-construction has been much studied in Mandarin, the same in not true for Cantonese. Although this construction has traditionally been termed a "passive", I will show that it can have a different range of semantic interpretations in Cantonese. I argue that BEI is not confined to passive, but is used under certain circumstances to form a causative construction as well. The differences in behaviour between passive-BEI and causative-BEI can be seen in tests with anaphoric binding. I conclude that while the passive structure is mono-clausal, the causative structure must be bi-clausal. The Cantonese BEI-constructions have an obligatory agent-phrase which cannot be dropped. This differs from Mandarin and the challenge is to find an account for this phenomenon, especially if we are to claim that this construction is a passive. The optionality of the agent phrase is characteristic of passives and yet Cantonese deviates from this norm. I argue that passive in Cantonese is a syntactic process and predict that only transitive verbs may participate in this construction. I utilize the universal v-VP structure on transitive verbs, proposed by Chomsky (1995), to guarantee that the external theta role must be retained. I also examine the much debated status of BEI which is used in the BEI-construction. Although this construction can be used to derive both a passives and a causatives, it does not necessarily mean that two separate BEIs must be posited. I conclude that BEI can be treated as a category-neutral element which can interact in both causative and passive structures. To support this proposal I appeal to the functional versus lexical distinction of categories and projections.
This paper proposes a novel method to refine the grammars in parsing by utilizing semantic knowledge from HowNet. Based on the hierarchical state-split approach, which can refine grammars automatically in a data-driven manner, this study introduces semantic knowledge into the splitting process at two steps. Firstly, each part-of-speech node will be annotated with a semantic tag of its terminal word. These new tags generated in this step are semantic-related, which can provide a good start for splitting. Secondly, a knowledge-based criterion is used to supervise the hierarchical splitting of these semantic-related tags, which can alleviate overfitting. The experiments are carried out on both Chinese and English Penn Treebank show that the refined grammars with semantic knowledge can improve parsing performance significantly. Especially with respect to Chinese, our parser achieves an F 1 score of 87.5%, which is the best published result we are aware of.