Papers reviewed and determined not to be word norm studies. Use the flag icon to report errors or suggest re-inclusion.
16504 papers
We describe a model for the lexical analy-sis of Arabic text, using the lists of alterna-tives supplied by a broad-coverage morpho-logical analyzer, SAMA, which include sta-ble lemma IDs that correspond to combina-tions of broad word sense categories and POS tags. We break down each of the hundreds of thousands of possible lexical labels into its constituent elements, including lemma ID and part-of-speech. Features are computed for each lexical token based on its local and document-level context and used in a novel, simple, and highly efficient two-stage super-vised machine learning algorithm that over-comes the extreme sparsity of label distribu-tion in the training data. The resulting system achieves accuracy of 90.6 % for its first choice, and 96.2 % for its top two choices, in selecting among the alternatives provided by the SAMA lexical analyzer. We have successfully used this system in applications such as an online reading helper for intermediate learners of the Arabic language, and a tool for improving the productivity of Arabic Treebank annotators. 1
Understanding the syntactic structure of a sentence is a necessary preliminary to understanding its semantics and therefore for many practical applications. The field of natural language processing has achieved a high degree of accuracy in parsing, at least in English. However, the syntactic structures produced by the most commonly used parsers are less detailed than those structures found in the treebanks the parsers were trained on. In particular, these parsers typically lack the null elements used to indicate wh-movement, control, and other phenomena. This thesis presents a system for inserting these null elements into parse trees in English. It then examines the problem in Arabic, which motivates a second, joint-inference system which has improved performance on English as well. Finally, it examines the application of information derived from the Google Web 1T corpus as a way of reducing certain data sparsity issues related to wh-movement.
In this paper, we propose automatic categorization and summarization of documentaries using subtitles of videos. We propose two methods for video categorization. The first makes unsupervised categorization by applying natural language processing techniques on video subtitles and uses the WordNet lexical database and WordNet domains. The second has the same extraction steps but uses a learning module to categorize. Experiments with documentary videos give promising results in discovering the correct categories of videos. We also propose a video summarization method using the subtitles of videos and text summarization techniques. Significant sentences in the subtitles of a video are identified using these techniques and a video summary is then composed by finding the video parts corresponding to these summary sentences.
Proceedings of the Ninth International Workshop \non Treebanks and Linguistic Theories. \nEditors: Markus Dickinson, Kaili Müürisep and Marco Passarotti. \nNEALT Proceedings Series, Vol. 9 (2010), 19-30. \n© 2010 The editors and contributors. \nPublished by \nNorthern European Association for Language \nTechnology (NEALT) \nhttp://omilia.uio.no/nealt. \nElectronically published at \nTartu University Library (Estonia) \nhttp://hdl.handle.net/10062/15891.
This paper investigates whether high-quality annotations for tasks involving semantic disambiguation can be obtained without a major investment in time or expense. We examine the use of untrained human volunteers from Amazon’s Mechanical Turk in disambiguating prepositional phrase (PP) attachment over sentences drawn from the Wall Street Journal corpus. Our goal is to compare the performance of these crowdsourced judgments to the annotations supplied by trained linguists for the Penn Treebank project in order to indicate the viability of this approach for annotation projects that involve contextual disambiguation. The results of our experiments show that invoking majority agreement between multiple human workers can yield PP attachments with fairly high precision, confirming that this crowdsourcing approach to syntactic annotation holds promise for the generation of training corpora in new domains and genres.
Proceedings of the Ninth International Workshop \non Treebanks and Linguistic Theories. \nEditors: Markus Dickinson, Kaili Müürisep and Marco Passarotti. \nNEALT Proceedings Series, Vol. 9 (2010), 67-78. \n© 2010 The editors and contributors. \nPublished by \nNorthern European Association for Language \nTechnology (NEALT) \nhttp://omilia.uio.no/nealt. \nElectronically published at \nTartu University Library (Estonia) \nhttp://hdl.handle.net/10062/15891.
We investigate Arabic Context Free Grammar parsing with dependency annotation comparing lexicalised and unlexicalised parsers. We study how morphosyntactic as well as function tag information percolation in the form of grammar transforms (Johnson, 1998, Kulick et al., 2006) affects the performance of a parser and helps dependency assignment. We focus on the three most frequent functional\ntags in the Arabic Penn Treebank: subjects, direct objects and predicates. We merge these functional tags with their phrasal categories and (where appropriate) percolate case information to the non-terminal (POS) category to train the parsers. We then automatically enrich the output of these parsers with full dependency information in order to annotate trees with Lexical Functional Grammar (LFG)\nf-structure equations with produce f-structures, i.e. attribute-value matrices approximating to basic predicate-argument-adjunct structure representations. We present a series of experiments evaluating how well lexicalized, history-based, generative (Bikel) as well as latent\nvariable PCFG (Berkeley) parsers cope with the enriched Arabic data. We measure quality and coverage of both the output trees and the generated LFG f-structures. We show that joint functional and morphological information percolation improves both the recovery of trees as well as dependency results in the form of LFG f-structures.
Real world audio clips contain numerous acoustic sources. The rich acoustic information they carry cannot be fully described with single or even multiple terms about the acoustic sources alone. For instance, the label birds assigned to a birds singing clip that includes sounds of trees and a small river does not properly capture the experience it creates in the person listening to it. In this paper we introduce a novel scheme where the subjective experience of listening to sound clips containing mixture of sources is captured using affective measures. Furthermore, in contrast to the conventional approach of simple label-based methods, the affective ratings are then used to evaluate the performance of an example-based audio retrieval system. We argue that audio retrieval systems can benefit from using affective measures which are well established in experimental psychology, especially when dealing with real world audio clips. We present experimental results of our pilot study to support this motivation where the latent indexing framework has been employed for example-based retrieval on a collection of clips from the BBC sound effects library. The result of our study indicates that using the scheme of affective measures for representation and evaluation is indeed a promising direction to explore.
BACKGROUND: While research indicates cognitive biases in individuals vulnerable to bipolar symptoms, the specificity of these biases to the self and others, and to low or high activation states, is underexplored. METHOD: These biases were investigated using individuals with high (n = 24) and low levels (n = 24) of hypomanic personality (HPS) during word rating and free recall of a list of trait words after a positive versus neutral mood induction. RESULTS: The mood induction was not successful. Also, in contrast to the predicted self-serving bias, there was a self-denigratory bias in self-ratings relative to ratings of another person. In post hoc analyses, the study succeeded in producing a rating task of trait words that differentiated between high and low hypomania-prone individuals, as the high HPS group made higher ratings of high activation trait words (e.g. dynamic) to describe positive and neutral attributes regardless of whether it referred to themselves or another person. The high HPS group also showed a negative recall bias, but it was not specific to the self, questioning assumptions made about negative biases in existing research. Furthermore, a strong relationship emerged between greater use of imagery at encoding and greater recall of self-referent, positive, high activation words, suggesting a role for the intensity of images associated with the amplification of emotions in people with bipolar disorder (Holmes et al., 2008). CONCLUSIONS: It seems important to consider various multiple factors in memory bias research in people vulnerable to bipolar disorder, including self versus other encoding, high versus low activated states and the role of mental imagery. Further research is needed to spell out their interactive contribution.
Syntactic consistency is the preference to reuse a syntactic construction shortly after its appearance in a discourse. We present an analysis of the WSJ portion of the Penn Tree-bank, and show that syntactic consistency is pervasive across productions with various left-hand side nonterminals. Then, we implement a reranking constituent parser that makes use of extra-sentential context in its feature set. Using a linear-chain conditional random field, we improve parsing accuracy over the generative baseline parser on the Penn Treebank WSJ corpus, rivalling a similar model that does not make use of context. We show that the context-aware and the context-ignorant rerankers perform well on different subsets of the evaluation data, suggesting a combined approach would provide further improvement. We also compare parses made by models, and suggest that context can be useful for parsing by capturing structural dependencies between sentences as opposed to lexically governed dependencies.
Proceedings of the Ninth International Workshop \non Treebanks and Linguistic Theories. \nEditors: Markus Dickinson, Kaili Müürisep and Marco Passarotti. \nNEALT Proceedings Series, Vol. 9 (2010), 163-174. \n© 2010 The editors and contributors. \nPublished by \nNorthern European Association for Language \nTechnology (NEALT) \nhttp://omilia.uio.no/nealt. \nElectronically published at \nTartu University Library (Estonia) \nhttp://hdl.handle.net/10062/15891.
DeSR is a statistical transition-based dependency parser which learns from annotated corpora which actions to perform for building parse trees while scanning a sentence. We describe the experiments performed for the ICON 2010 Tools Contest on Indian Dependency Parsing. DesR was configured to exploit specific features from the Indian treebanks. The submitted run used a stacked combination of four configurations of the DeSR parser and achieved the best unlabeled accuracy scores in all languages. The contribution to the result of various choices is analyzed.
This paper presents a method for the automatic detection and correction of malapropism errors found in documents using the WordNet lexical database, a search engine (Google) and a paronyms dictionary. The malapropisms detection is based on the evaluation of the cohesion of the local context using the search engine, while the correction is done using the whole text cohesion evaluated in terms of lexical chains built using the linguistic ontology. The correction candidates, which are taken from the paronyms dictionary, are evaluated versus the local and the whole text cohesion in order to find the best candidate that is chosen for replacement. The testing methods of the application are presented, along with the obtained results.
We propose an effective approach to automatically identify predicate heads in Chinese sentences based on statistical pre-processing and rule-based post-processing. In the preprocessing stage, the maximal noun phrases in a sentence are recognized and replaced by “NP ” labels to simplify the sentence structure. Then a CRF model is trained to recognize the predicate heads of this simplified sentence. In the post-processing stage, a rule base is built according to the grammatical features of predicate heads. It is then utilized to correct the preliminary recognition results. Experimental results show that our approach is feasible and effective, and its accuracy achieves 89.14 % on Tsinghua Chinese Treebank. 1
In this paper, we discuss our analysis and resulting new annotations of Penn Discourse Treebank (PDTB) data tagged as Concession. Concession arises whenever one of the two arguments creates an expectation, and the other ones denies it. In Natural Languages, typical discourse connectives conveying Concession are ‘but’, ‘although’, ‘nevertheless’, etc. Extending previous theoretical accounts, our corpus analysis reveals that concessive interpretations are due to different sources of expectation, each giving rise to critical inferences about the relationship of the involved eventualities. We identify four different sources of expectation: Causality, Implication, Correlation, and Implicature. The reliability of these categories is supported by a high inter-annotator agreement score, computed over a sample of one thousand tokens of explicit connectives annotated as Concession in PDTB. Following earlier work of (Hobbs, 1998) and (Davidson, 1967) notion of reification, we extend the logical account of Concession originally proposed in (Robaldo et al., 2008) to provide refined formal descriptions for the first three mentioned sources of expectations in Concessive relations. 1.
Towards human consistent data driven decision support systems using verbalization of data mining results via linguistic data summaries We present how the conceptually and numerically simple concept of a fuzzy linguistic database summary can be a very powerful tool for gaining much insight into the essence of data that may be relevant for a business activity. The use of linguistic summaries provides tools for the verbalization of data analysis (mining) results which, in addition to the more commonly used visualization e.g. via a GUI, graphical user interface, can contribute to an increased human consistency and ease of use. The results (knowledge) derived are in a simple, easily comprehensible linguistic form which can be effectively and efficiently employed for supporting decision makers via the data driven decision support system paradigm. Two new relevant aspects of the analysis are also outlined which was first initiated by the authors. First, following Kacprzyk and Zadrożny [1] comments are given on an extremely relevant aspect of scalability of linguistic summarization of data, using their new concept of a conceptual scalability that is crucial for large applications. Second, following Kacprzyk and Zadrożny [2] it is further considered how linguistic data summarization is closely related to some types of solutions used in natural language generation (NLG), which can make it possible to use more and more effective and efficient tools and techniques developed in this another rapidly developing area. An application of a computer retailer is outlined.
In this paper authors will show how verb valency data, added to the Croatian dictionary in NooJ, enhances recognition of VP as well as NP and PP parts of a sentence. At the Department of Information Sciences two parallel PhD Theses were being developed. One is construction of a chunker for Croatian using NooJ and the other one is construction of Croatian verb valency lexicon (CRVLLEX). Here we combined the two projects by adding the data from CRVLLEX to the existing NooJ dictionary hoping thus to obtain better, improved Croatian chunker results. Our dictionary has over 36 000 entries of which 1 884 are verbs. Each verb is only marked by its category and the FLX. Additional data is being added from the CRVLLEX. Theoretic motivation behind the construction of CRVLLEX is Praha’s Dependency Treebank (PDT) but also good Czech verb valency dictionaries (VALLEX and Verbalex). So far, CRVLLEX has 1 739 verbs with 5 118 valency frames (approx. 3 frames per verb) and 173 syntactic-semantic classes. Each word entry in CRVLLEX contains headword lemma, reflexivity example (depending whether the verb is reflexive or not) and frame entry which is the part that describes the valency frame for each verb. Every verb also has the following attributes: aspect, frequency and form, which we hope will be of great importance in disambiguating parts of a sentence surrounding the verb.
From the perspective of translation, Lestrygonians is a chapter that goes in the face of linguistic norm, subverting expectations of standard language use. The chapter consistently shows meaning to be unstable, self-generating along lines of flight, mushrooming around nodes of (semantic) tension. The cohesion of the text, as shown in the sandwich’ passage, depends on a series of such nodes, of linguistic occasions, associations and interfaces; Bloom’s associations and bricolage (or, to use Jean- Michel Rabaté’s pun, brick-holing)1 of words which direct and channel the text’s movements are called forth by the shape of phrases, homonymy and polysemy. Such instability of meaning and of the directions of the production of meaning certainly musters and breeds translation problems: the translator would have to aim at finding/ creating analogous interfaces in the target language (TL) that could account for, and function as points of departure for Bloom’s strings of associations, for the text’s nodes of meaning and tension.
We present a new reference Grammar of French (La Grande Grammaire du français), which is a collective project (gathering around fifty contributors), producing a book (about 2200 pages, to be published en 2011) and associated databases. Like the recent reference grammars of the other Romance Languages, it takes into account the important results of the linguistic research of the past thrity years, while aiming at a non specialist audience and avoiding formalization. We differ from existing French grammar by being focused on contemporary French from a purely descriptive point of view, and by taking spoken data into account. We include a description of all the syntactic phenomena, as well as lexical, semantic, pragmatic and prosodic insights, specially as they interact with syntax. The analysis concerns the data from contemporary written French, but also includes data from spoken corpora and regional or non standard French (when accessible). Throughout the grammar, a simple phrase structure grammar is used, in order to maintain a common representation. The analyses are modular with a strict division of labor between morphology, syntax and semantics. From the syntactic point of view, POS are also distinguished from grammatical relations (or functions). The databases include a terminological glossary, different lexical databases for certain POS, certain valence frames and certain semantic classes, and a bibliographical database. 1. Presentation of the project Sizable comprehensive grammars, written by linguists but aiming at a larger audience, have been published for several European languages: for Italian, the Grande
Linguistic variants are elements that are considered identically in terms of functionality, though they differ formally. The present article discusses variant forms of nouns. The study refers both to variants in the strict sense, i.e. identical in terms of function, and oscillating variants, i.e. those that differ in one quality, i.e. chronology, frequency, social range, stylistic character or the degree of expressiveness or emotiveness. The examples have been excerpted from Nowy słownik poprawnej polszczyzny PWN [The PWN New Dictionary of Correct Polish Usage]. The collected material has been classified into several groups depending on the relations between variants defined by means of qualifiers and abbreviations. Formal and linguistic expressions of noun variantivity are also presented. The examples under scrutiny indicate that the linguistic norm (standard) accepts far more nominal variant forms than the linguistic custom presents as it has sanctioned only a certain part of the analysed examples. A detailed and thorough study on the functioning of nominal variants in the linguistic custom is, however, difficult to carry out.
A layer-based projective dependency parsing approach is presented.This novel approach works layer by layer in a bottom-up manner,in which the depth of token dependency is allowed no more than one.Inside the layer the dependency graphs are searched exhaustively while between the layers the parser state transfers deterministically.Taking the dependency layer as the parsing unit,the proposed parser has a lower computational complexity than graph-based models which search for a whole dependency graph,alleviating the error propagation in transition-based models to some extent.Furthermore,our parser adopts the sequence labeling models to find the optimal sub-graph of the layer,which demonstrates the sequence labeling techniques qualified for hierarchical structure analysis tasks.Experimental results indicate that the proposed approach offers desirable accuracies and especially a very fast parsing speed,with 2500 words per second for Penn Treebank.
Whereas dictionary design has traditionally been guided by the results of dictionary use research, recent approaches in lexicographic research are strictly user-centred. We support the idea of integrating empirical cognitive evidence into this type of research, thus fruitfully exploiting it for both, the selection (and subsequently presentation) of lexical data and the acquisition of such data from corpora. Focusing on the extraction of semantic relations to be illustrated in electronic learners ’ dictionaries, we analyse the results of two behavioural experiments on the production as well as the perception of semantic relations. The main goal of the experiments was to determine which relations are cognitively salient in speakers ’ minds. With the objective of developing a method to automatically extract cognitively salient semantic relations from corpora, we describe and discuss findings of the first analyses conducted on composite part relations. In future this might serve as a basis for the elaboration of new strategies aimed at enriching lexical databases and dictionaries.
Do you ever complete, select, write, or administer rating scales to provide data for decision making? NO? Then you must be working in a storm drain. Most faculty developers I know have a wide variety of rating scales that fly across their desk tops as their incremental program activities unfold during the academic year. The primary issue for this column is: What is the quality of those ratings used for decisions about people and programs? When students, faculty, and administrators rate a program or someone's performance, we assume that they will read each item carefully and make their honest assessment with scrupulous impartiality (LOL). Are we dreaming in 3D IMAX like Avatar? The problem is that human tendencies may contaminate their responses, rendering them less than honest and impartial. In fact, these tendencies may be driven by demonic forces, such as those seen on Fringe or The Vampire Diaries or appearing in Stephen King's books. However, they may be conscious and intentional or unconscious. These tendencies are known as response sets or biases. More than 10 types of bias can affect ratings, such as halo effect; end-aversion bias; extreme-response bias; acquiescence bias; and gender, racial/ ethnic, and sexual orientation bias. There are four that particularly afflict faculty or administrator ratings: leniency-stringency bias, incompetence bias, buddy bias, and back-scratching bias. Your awareness of these biases and the techniques to minimize or eliminate them can produce more accurate ratings and foster more perceptive decisions than those currently being made. These biases are briefly described next along with suggestions for minimizing their effects. For a more detailed, rollicking treatment of the topic with research citations, see Berk (2006), Nunnally and Bernstein (1994), and Streiner and Norman (1995). 1. Halo Effect This effect is the extent to which a rater's overall impression of a person will affect his or her rating on each item. For example, if the global impression is positive and a student really likes her instructor, she may simply mark Strongly Agree to all positive statements. However, despite the positive image of a halo on an angel (Remember: The TV series Touched by a Halo?), this effect can also be negative, such as when a student hates the course. Possible solutions: Since this is an individual bias, the larger the group of raters, the greater the chance that positive and negative halo effects will cancel each other out. Unfortunately, there is no simple solution to control or minimize halo effect. 2. End-Aversion/Central Tendency Bias This is the tendency to ignore the extreme response options or anchors on the scale because they may be viewed as too strong. When the extremes are not selected, the ratings may be squished into the middle of the scale, which restricts the range of responses. Possible solutions: This bias should not be a problem on most rating scales in the agree-disagree format. However, scales consisting of several evaluation anchors, such as excellent-poor, or frequency anchors with absolutes, such as always-never, can produce this type of bias. Either soften the extreme anchors with Almost Always and Almost Never, or extend the number of anchors with the expectation that the extremes will rarely be chosen. 3. Extreme-Response Bias This is the opposite of end-aversion bias. In this case, the respondents may mark the extreme anchors rather than those in between. This bias is difficult to detect because the reason for the choice of the extremes may also be due to their honest ratings or the halo effect. Possible solution: This type of bias may occur less often than the other sources of bias. However, there is no direct solution. 4. Acquiescence/Affirmation/ Agreement/Yea-Saying Bias This is the tendency to agree or give positive responses to statements irrespective of their content. …
1 Many researchers have explored the topic with the intent of identifying the logic of the variation between CC and non-CC. In this paper we want to address a specific problem pointed out in the literature about CC in Spanish: its incidence is far less frequent in written texts. This can be interpreted as a problem in face of the diachronic evolution of the phenomenon, which showed higher frequency in medieval times. Therefore, the tendency for grammatical innovations (non-CC) to be more common in speech seems to be contradicted. I want to shed light on this problem by assigning the difference between registers to the dependence of the phenomenon to Information Structure. In order to do so, I demonstrate not only that register is a relevant variation factor in European Portuguese (EP), but also that it is connected to internal factors which are relevant to understand variation. The paper presents the results from a EP corpus made up of about 1,000 data from different registers and sources: we have collected sentences from interviews given to linguistic databases and newspapers through the Corpus do Portugues (Davies & Ferreira 2006), and from contemporary Portuguese novels, all listed in the Appendix. The sentences include a finite verb followed by an infinitive, once CC is obligatory in EP in verb sequences involving a gerund or a participle. Before moving along, it is necessary to differentiate between two constructions that show CC: and union. 2 While restructuring is usually linked to governing verbs of the modal, aspectual and movement classes which usually display semantic bleaching - see (1) - in union the governing verb is causative or perceptive. A verb of the latter type selects a complement with a non-coreferent subject. This embedded subject occurs in accusative or dative form, and climbs obligatorily, alongside other clitics - (2)b. Once there is no variation with respect to CC in clause union, we will only deal with restructuring data, similarly to previous studies about variation in CC.
The past several years have witnessed rapid advances in syntax-based machine translation, which exploits natural language syntax to guide translation. Depending on the type of input, most of these efforts can be divided into two broad categories: (a) string-based systems whose input is a string, which is simultaneously parsed and translated by a synchronous grammar (Wu, 1997; Chiang, 2005; Galley et al., 2006), and (b) tree-based systems whose input is already a parse tree to be directly converted into a target tree or string (Lin, 2004; Ding and Palmer, 2005; Quirk et al., 2005; Liu et al., 2006; Huang et al., 2006). Compared with their string-based counterparts, tree-based systems offer many attractive features: they are much faster in decoding (linear time vs. cubic time), do not require sophisticated binarization (Zhang et al., 2006), and can use separate grammars for parsing and translation (e.g. a context-free grammar for the former and a tree substitution grammar for the latter). However, despite these advantages, most treebased systems suffer from a major drawback: they only use 1-best parse trees to direct translation, which potentially introduces translation mistakes due to parsing errors (Quirk and Corston-Oliver, 2006). This situation becomes worse for resourcepoor source languages without enough Treebank data to train a high-accuracy parser. This problem can be alleviated elegantly by using packed forests (Huang, 2008), which encodes exponentially many parse trees in a polynomial space. Forest-based systems (Mi et al., 2008; Mi and Huang, 2008) thus take a packed forest instead of a parse tree as an input. In addition, packed forests could also be used for translation rule extraction, which helps alleviate the propagation of parsing errors into rule set. Forest-based translation can be regarded as a compromise between the string-based and tree-based methods, while combining the advantages of both: decoding is still fast, yet does not commit to a single parse. Surprisingly, translating a forest of millions of trees is even faster than translating 30 individual trees, and offers significantly better translation quality. This approach has since become a popular topic.
Human communication takes place in both spoken and written modes, as in face-to-face conversations, books reading, television watching and Internet browsing. However, for people with language disabilities (e.g. aphasia), people with low literacy, and people with poor command of a language, receiving and expressing information via a language is difficult. Because of the inability to comprehend words and/or to find words that express intended concepts, people with language barriers may encounter great challenges performing everyday tasks such as ordering food and visiting a doctor. As an alternative to words, pictures and sounds have been designed, tested, and used to evoke concepts in computer-based interfaces, education, industry, and advertisement. However, icons created by artists, user-uploaded photos, auditory icons and earcons cannot always satisfy the need for communicating everyday concepts for people with language barriers. Furthermore, the iconic vocabularies in current communication support systems have ambiguous representations and difficult to use vocabulary organization. My thesis research addresses issues in the above paragraph and explores building multimedia support for information delivery. The work is unique in that it provides different channels (visual and auditory) to support nonverbal communication, and that it has a disambiguated sense-level network structure with semantic links between related concepts. The development of such a multimedia-enhanced lexical database is challenging, due to the lack of guidance in the creation and evaluation of multimodal representations. To construct the lexical database, I went through a cycle of design, construction, evaluation, and modification for each media modality-to-concept association. I have introduced novel representations that are under investigated (e.g. videos and environmental sounds). I have explored difference sources of multimedia data (designed by specialists, designed by general users, and web resources not tailored for communication purposes). I have experimented with new methodologies for generating and evaluating multimedia-concept associations. I have also come up with essential guidelines for better multimedia vocabulary design. A communication support system, the Online Multimedia Language Assistant (OMLA), which employs various multimedia forms, including web images, icons, animations, videos, and environmental sounds, was built on top of my research on multimedia representations with the goal of assisting comprehension of common concepts across language barriers. OMLA is implemented as a popup dictionary in the form of a web browser extension. Users can select an unfamiliar word on a webpage to view its associated visual/auditory representation in a popup box. The Language Assistant can enhance concept understanding as people browse information on the Internet, and support face-to-face communication when people want to illustrate a term via a picture, a video or a sound. Through a series of studies, I verified that web images are as effective as stylized icons in conveying nouns, that videos outperformed other stimuli in illustrating verbs, and that nonspeech sounds are better in distinguishing many concepts like thunder, alarms, and sneezing. The Multimedia Augmented Online Language Assistant was shown to enhance information comprehension, and I am exploring its application in real life scenarios such as medical care.
Building up messages as a cognitive activity within the linguistic multi-level system is the result of the interaction between the various components of this system. Yet, this interactive process occurring in the language user's mind while encoding can vary from person to person. Likewise, it also differs in different recipients while decoding. This cognitively carried out difference in encoding can result in either unintended messages, though emptied out in mature and well-formed language constructions, or ill-formed ones. The present study aims at identifying the most frequent syntactic errors at the sentential level; and how immature or vague conceptualization manifests itself in the grammar-meaning relationship as reflected in the subjects' errors. Data were collected from the writings and the performance tests of mainly two Arab groups of third and fourth-year English majors in the course of Sociolinguistics at the University of Nizwa in the Sultanate of Oman. Errors were identified, classified and interpreted in terms of the underlying cognitive processes they went through during production. Errors were related to time-tense vague mapping, finite-nonfinite confusion, sentence-clause uncertainty, incorrect embedding, voice-related inaccuracy and verbless clauses or sentences. ********** Errors in second language performance are easy to observe, and undoubtedly serve as good indicators of a learner's level of second-language knowledge. Thus, they are an integral component of language learning. The phenomenon of error has received the interest of researchers, though traditionally has been regarded as the linguistic phenomena deviant from the language rules and standard usages, reflecting learners' deficiency in language competence. Hence, many teachers simply correct individual errors as they occur, giving little concern to identifying patterns of errors or to uncovering causes other than learner ignorance. However, interpreting what the cause of these errors has been a major concern for linguists, as well as classroom practitioners. In this respect, Steinberg and Sciarini (2006), state that the systematicity of most errors is the result of the application of certain strategies when relevant second-language knowledge is not available or incomplete. Conscious resort to any of these strategies is definitely a cognitive exercise, the results of which are various deviations of the linguistic norms, sentence patterns and rules that put the language components together to produce linguistic moulds embodying composites of thoughts. Presently, therefore, with the development of linguistics, applied linguistics, psychology and other relevant subjects, attitudes toward errors changed greatly. Instead of being problems to be overcome, errors are viewed as an evidence of the learners' stages in their target language (TL) development. It is through analyzing learner errors that errors are elevated from the status of undesirability to that of a guide to the inner working of the language learning process, as (Ellis, 1985, cited by Nassaji & Fotos, 2004, at http://faculty.ksu.edu.sa/dinaalsibai/ReviewGrammar.PDF) puts it. The term grammar is used in this paper to include the two language levels of syntax and morphology since some errors in one level cannot be explained in isolation from the other. For example, the past tense of bring is brought, a case of inflection under morphology, whereas the interrogative form Did bring them? of Ali brought them is a syntactic issue. Memory and logic, which are cognitive skills, play a crucial role in language learning. As Steinberg and Sciarini (2006: 34-35) state it, that while learning to identify the words of the language, formulating rules for their use, and relating speech to the environment and mind, the child employs an exceptional memory capacity. In this respect, a massive number of particular words, phrases and sentences have to be remembered in connection with the context, whether physical or mental, in which they took place. …
Introduzione L’impiego di corpora elettronici nella ricerca linguistica si rivela partico larmente efficace quando ai dati raccolti ( raw data ) vengono aggiunte informazioni di carattere linguistico e funzionale che facilitano la loro in terpretazione da un punto di vista morfosintattico. Il risultato di tali pro cedure di analisi e rappresentato da Treebanks ovvero corpora annotati sintatticamente attraverso rappresentazioni ad albero. Solitamente distinta dalla procedura di mark-up , l’annotazione di un corpus realizzata con l’applicazione di sistemi computerizzati ( parsers ) che eseguono un’etichettatura delle parti del discorso ( POS tagging ) costituisce una procedura particolarmente efficace nel chiarire, ad esempio, casi di ambiguita grammaticale in cui un lessema, se privo di annotazione, puo essere interpretato sia come verbo (V) che come nome (N) (ad esempio work ). I parsers generalmente propongono descrizioni delle produzioni linguistiche di una comunita di parlanti basate su regole grammaticali estratte da un campione di dati e tale caratteristica talvolta non permette di eseguire l’analisi automatica anche di frasi non grammaticali . Si tratta, dunque, di sistemi di analisi basati su grammatiche di precisione che di stinguono principalmente cio che e grammaticale da cio che non lo e ba sandosi su un corpus necessariamente limitato di dati, in quanto la descri zione fa esclusivamente riferimento a campioni di frasi grammaticalmen te corrette di una lingua inseriti nel sistema. Tuttavia, la propensione dei parlanti a produrre costruzioni che presentano “deviazioni” dalla norma e il fatto che molti di essi usino una lingua straniera come strumento di co municazione comporta in se «that naturalistic ungrammatical sentences are of interest to linguists studying language production, language loss and language learning, and that the grammatical/ungrammatical distinc tion cannot therefore be completely dismissed» .
This study investigates how the duration of the stem vowel of regular and irregular English verbs is modulated by tense (present and past), regularity, lexical frequency, gang size of the vocalic alternation, imageability ratings, and vowel quality. The vocalic durations of 48 monosyllabic irregular verbs and 171 regular verbs were extracted from the Buckeye Corpus of spontaneous speech. A linear mixed effects regression model revealed that vowels of past tense forms tend to have longer durations than vowels of present tense forms, that vowels of words that are less imageable are realized with shorter durations, and that tense vowels are longer than lax vowels. Surprisingly, higher frequency irregular past tense forms were produced with longer vowels, contradicting Aylett and Turk [(2004); (2006)] and Bell et al. [(2003); (2009)]. Further, vowels were shorter for irregular verbs with larger vocalic alternation gangs, contradicting the predictions of Kuperman et al. [(2007)] but supporting hypotheses that units with a smaller information load have shorter durations. This pattern of results is interpreted as a consequence of the pressure for regularization during the production of irregular past tense forms.
Motion pictures are so typical experience goods that consumers tend to look for more credible information. Hence, movie audiences consider movie viewers` reviews more important than the information provided by the film distributor. Recently many portal sites allow consumers to post their reviews and opinions so that other people check the number of consumer reviews and scores before going to the theater. There are a few previous researches studying the electronic word of mouth(eWOM) effect in the movie industry. They found that the volume of eWOM influenced the revenue of the movie significantly but the valence of eWOM did not affect it much (Liu 2006). The goal of our research is also to investigate the eWOM effects in general. But our research is different from the previous studies in several aspects. First, we study the eWOM effect in Korean movie industry. In other words, we would like to check whether we can generalize the results of the previous research across countries. The similar econometric models are applied to Korean movie data that include 746,282 consumer reviews on 439 movies. Our results show that both the valence(RATING) and the volume(LNMSG) of the eWOM influence weekly movie revenues. This result is different from the previous research findings that the volume only influences the revenue. We conjectured that the difference of self construal between Asian and American culture may explain this difference (Kitayama 1991). Asians including Koreans have more interdependent self construal than American, so that they are easily affected by other people`s thought and suggestion. Hence, the valence of the eWOM affects Koreans` choice of the movie. Second, we find the critical defect of the previous eWOM models and, hence, attempt to correct it. The previous eWOM model assumes that the volume of eWOM (LNMSG) is an independent variable affecting the movie revenue (LNREV). However, the revenue can influence the volume of the eWOM. We think that treating the volume of eWOM as an independent variable a priori is too restrictive. In order to remedy this problem, we employed a simultaneous equation in which the movie revenue and the volume of the eWOM can affect each other. That is, our eWOM model assumes that the revenue (LNREV) and the volume of eWOM (LNMSG) have endogenous relationship where they influence each other. The results from this simultaneous equation model showed that the movie revenue and the eWOM volume interact each other. The movie revenue influences the eWOM volume for the entire 8 weeks. The reverse effect is more complex. Both the volume and the valence of eWOM affect the revenue in the first week, but only the volume affect the revenue for the rest of the weeks. In the first week, consumers may be curious about the movie and look for various kinds of information they can trust, so that they use the both the quantity and quality of consumer reviews. But from the second week, the quality of the eWOM only affects the movie revenue, implying that the review ratings are more important than the number of reviews. Third, our results show that the ratings by professional critics (CRATING) had negative effect to the weekly movie revenue (LNREV). Professional critics often give low ratings to the blockbuster movies that do not have much cinematic quality. Experienced audiences who watch the movie for fun do not trust the professionals` ratings and, hence, tend to go for the low-rated movies by them. In summary, applied to the Korean movie ratings data and employing a simultaneous model, our results are different from the previous eWOM studies: 1) Koreans (or Asians) care about the others` evaluation quality more than quantity, 2) The volume of eWOM is not the cause but the result of the revenue, 3) Professional reviews can give the negative effect to the movie revenue.
Natural language processing has several ambiguity problems, and English-Korean machine translation especially includes those problems to be solved in each translation step. This paper focuses on resolving part-of-speech ambiguity of English words in order to improve the efficiency of English analysis, which is in part of efforts for developing practical English-Korean machine translation system. In order to improve the efficiency of the English analysis, the part-of-speech determination must be fast and accurate for being integrated with machine translation system. This paper proposes the probabilistic models for part-of-speech determination. We use Penn Treebank corpus in building the probabilistic models. In experiment, we present the performance of the part-of-speech determination models and the efficiency improvement of the machine translation system by the proposed part-of-speech determination method.
Background: A crucial question for understanding sentence comprehension is the openness of syntactic and semantic processes for other sources of information. Using event-related potentials in a dual task paradigm, we had previously found that sentence processing takes into consideration task relevant sentence-external semantic but not syntactic information. In that study, internal and external information both varied within the same linguistic domain—either semantic or syntactic. Here we investigated whether across-domain sentence-external information would impact within-sentence processing. Methodology: In one condition, adjectives within visually presented sentences of the structure [Det]-[Noun]-[Adjective]- [Verb] were semantically correct or incorrect. Simultaneously with the noun, auditory adjectives were presented that morphosyntactically matched or mismatched the visual adjectives with respect to gender. Findings: As expected, semantic violations within the sentence elicited N4)
As our understanding of the basic processes underlying reading is growing, the key role played by attention in this process becomes evident. Two research topics are of particular interest in this domain: (1) it is still undetermined whether sustained attention affects lexical decision tasks; (2) the influence of attention on early visual processing (i.e., before orthographic or lexico-semantic processing stages) remains largely under-specified. Here we investigated early perceptual modulations by sustained attention using an ERP paradigm adapted from Thierry et al. [1]. Participants had to decide whether visual stimuli presented in pairs pertained to a pre-specified category (lexical categorization focus on word or pseudoword pairs). Depending on the lexical category of the first item of a pair, participants either needed to fully process the second item (hold condition) or could release their attention and make a decision without full processing of the second item (release condition))
Background: Languages differ greatly both in their syntactic and morphological systems and in the social environments in which they exist. We challenge the view that language grammars are unrelated to social environments in which they are learned and used. Methodology/Principal Findings: We conducted a statistical analysis of >2,000 languages using a combination of demographic sources and the World Atlas of Language Structures- a database of structural language properties. We found strong relationships between linguistic factors related to morphological complexity, and demographic/socio-historical factors such as the number of language users, geographic spread, and degree of language contact. The analyses suggest that languages spoken by large groups have simpler inflectional morphology than languages spoken by smaller groups as measured on a variety of factors such as case systems and complexity of conjugations. Additionally, languages spoken by large groups are much more likely to u)
What kind of mental objects are letters? Research on letter perception has mainly focussed on the visual properties of letters, showing that orthographic representations are abstract and size/shape invariant. But given that letters are, by definition, mappings between symbols and sounds, what is the role of sound in orthographic representation? We present two experiments suggesting that letters are fundamentally sound-based representations. To examine the role of sound in orthographic representation, we took advantage of the multiple scripts of Japanese. We show two types of evidence that if a Japanese word is presented in a script it never appears in, this presentation immediately activates the ("actual") visual word form of that lexical item. First, equal amounts of masked repetition priming are observed for full repetition and when the prime appears in an atypical script. Second, visual word form frequency affects neuromagnetic measures already at 100-130 ms whether the word is pre)
We recently used computational phylogenetic methods on lexical data to test between two scenarios for the peopling of the Pacific. Our analyses of lexical data supported a pulse-pause scenario of Pacific settlement in which the Austronesian speakers originated in Taiwan around 5,200 years ago and rapidly spread through the Pacific in a series of expansion pulses and settlement pauses. We claimed that there was high congruence between traditional language subgroups and those observed in the language phylogenies, and that the estimated age of the Austronesian expansion at 5,200 years ago was consistent with the archaeological evidence. However, the congruence between the language phylogenies and the evidence from historical linguistics was not quantitatively assessed using tree comparison metrics. The robustness of the divergence time estimates to different calibration points was also not investigated exhaustively. Here we address these limitations by using a systematic tree comparison )
Background: When two targets are presented in close temporal proximity amongst a rapid serial visual stream of distractors, a period of disrupted attention and attenuated awareness lasting 200-500 ms follows identification of the first target (T1). This phenomenon is known as the "attentional blink" (AB) and is generally attributed to a failure to consolidate information in visual short-term memory due to depleted or disrupted attentional resources. Previous research has shown that items presented during the AB that fail to reach conscious awareness are still processed to relatively high levels, including the level of meaning. For example, missed word stimuli have been shown to prime later targets that are closely associated words. Although these findings have been interpreted as evidence for semantic processing during the AB, closely associated words (e.g., day-night) may also rely on specific, well-worn, lexical associative links which enhance attention to the relevant target. Metho)
The verb google is intriguing for the study of morphology, loanwords, assimilation, language contrast and neologisms. We present data for it for nineteen languages from nine language families.
Background: Decoding of frequency-modulated (FM) sounds is essential for phoneme identification. This study investigates selectivity to FM direction in the human auditory system. Methodology/Principal Findings: Magnetoencephalography was recorded in 10 adults during a two-tone adaptation paradigm with a 200-ms interstimulus-interval. Stimuli were pairs of either same or different frequency modulation direction. To control that FM repetition effects cannot be accounted for by their on- and offset properties, we additionally assessed responses to pairs of unmodulated tones with either same or different frequency composition. For the FM sweeps, N1m event-related magnetic field components were found at 103 and 130 ms after onset of the first (S1) and second stimulus (S2), respectively. This was followed by a sustained component starting at about 200 ms after S2. The sustained response was significantly stronger for stimulation with the same compared to different FM direction. This effect )
Background: Vision provides the most salient information with regard to stimulus motion, but audition can also provide important cues that affect visual motion perception. Here, we show that sounds containing no motion or positional cues can induce illusory visual motion perception for static visual objects. Methodology/Principal Findings: Two circles placed side by side were presented in alternation producing apparent motion perception and each onset was accompanied by a tone burst of a specific and unique frequency. After exposure to this visual apparent motion with tones for a few minutes, the tones became drivers for illusory motion perception. When the flash onset was synchronized to tones of alternating frequencies, a circle blinking at a fixed location was perceived as lateral motion in the same direction as the previously exposed apparent motion. Furthermore, the effect lasted at least for a few days. The effect was well observed at the retinal position that was previously exp)
Variation is a ubiquitous feature of speech. Listeners must take into account context-induced variation to recover the interlocutor's intended message. When listeners fail to normalize for context-induced variation properly, deviant percepts become seeds for new perceptual and production norms. In question is how deviant percepts accumulate in a systematic fashion to give rise to sound change (i.e., new pronunciation norms) within a given speech community. The present study investigated subjects' classification of /s/ and /∫ / before /a/ or /u/ spoken by a male or a female voice. Building on modern cognitive theories of autism-spectrum condition, which see variation in autism-spectrum condition in terms of individual differences in cognitive processing style, we established a significant correlation between individuals' normalization for phonetic context (i.e., whether the following vowel is /a/ or /u/) and talker voice variation (i.e., whether the talker is male or female) in speech )
Background: Matrix-assisted laser desorption ionisation time of flight mass spectrometry (MALDI TOF-MS) allows the identification of most bacteria and an increasing number of fungi. The potential for the highest clinical benefit of such methods would be in severe acute infections that require prompt treatment adapted to the infecting species. Our objective was to determine whether yeasts could be identified directly from a positive blood culture, avoiding the 1-3 days subculture step currently required before any therapeutic adjustments can be made. Methodology/Principal Findings: Using human blood spiked with Candida albicans to simulate blood cultures, we optimized protocols to obtain MALDI TOF-MS fingerprints where signals from blood proteins are reduced. Simulated cultures elaborated using a set of 12 strains belonging to 6 different species were then tested. Quantifiable spectral differences in the 5000-7400 Da mass range allowed to discriminate between these species and to build)
Discrete phonological phenomena form our conscious experience of language: continuous changes in pitch appear as distinct tones to the speakers of tone languages, whereas the speakers of quantity languages experience duration categorically. The categorical nature of our linguistic experience is directly reflected in the traditionally clear-cut linguistic classification of languages into tonal or non-tonal. However, some evidence suggests that duration and pitch are fundamentally interconnected and co-vary in signaling word meaning in non-tonal languages as well. We show that pitch information affects real-time language processing in a (non-tonal) quantity language. The results suggest that there is no unidirectional causal link from a genetically-based perceptual sensitivity towards pitch information to the appearance of a tone language. They further suggest that the contrastive categories tone and quantity may be based on simultaneously covarying properties of the speech signal and t)
Background: Absolute pitch (AP) is the ability to identify or produce isolated musical tones. It is evident primarily among individuals who started music lessons in early childhood. Because AP requires memory for specific pitches as well as learned associations with verbal labels (i.e., note names), it represents a unique opportunity to study interactions in memory between linguistic and nonlinguistic information. One untested hypothesis is that the pitch of voices may be difficult for AP possessors to identify. A musician's first instrument may also affect performance and extend the sensitive period for acquiring accurate AP. Methods/Principal Findings: A large sample of AP possessors was recruited on-line. Participants were required to identity test tones presented in four different timbres: piano, pure tone, natural (sung) voice, and synthesized voice. Note-naming accuracy was better for non-vocal (piano and pure tones) than for vocal (natural and synthesized voices) test tones. Th)
Korean Word Associations (KorWA) were collected to build a semantic network for the Korean language. A graphic representation approach of applying coefficients to complex networks allows us to discern the semantic structures within words. A semantic network of the KorWA was found to exhibit the scalefree property in its degree distribution. The growth of the network around hub words was also confirmed through two experimental phases. As an issue for further research, we suggest that the present results may yield insights for computational neurolinguistics, as a semantic network of word association norms can bridge the gap between information about lexical co-occurrences derived from a corpora and anatomical networks as a basis for mapping out neural activations. 1
Background: Prosody, the melody and intonation of speech, involves the rhythm, rate, pitch and voice quality to relay linguistic and emotional information from one individual to another. A significant component of human social communication depends upon interpreting and responding to another person's prosodic tone as well as one's own ability to produce prosodic speech. However there has been little work on whether the perception and production of prosody share common neural processes, and if so, how these might correlate with individual differences in social ability. Methods: The aim of the present study was to determine the degree to which perception and production of prosody rely on shared neural systems. Using fMRI, neural activity during perception and production of a meaningless phrase in different prosodic intonations was measured. Regions of overlap for production and perception of prosody were found in premotor regions, in particular the left inferior frontal gyrus (IFG). Act)
Background: The geographical position of Maharashtra state makes it rather essential to study the dispersal of modern humans in South Asia. Several hypotheses have been proposed to explain the cultural, linguistic and geographical affinity of the populations living in Maharashtra state with other South Asian populations. The genetic origin of populations living in this state is poorly understood and hitherto been described at low molecular resolution level. Methodology/Principal Findings: To address this issue, we have analyzed the mitochondrial DNA (mtDNA) of 185 individuals and NRY (non-recombining region of Y chromosome) of 98 individuals belonging to two major tribal populations of Maharashtra, and compared their molecular variations with that of 54 South Asian contemporary populations of adjacent states. Inter and intra population comparisons reveal that the maternal gene pool of Maharashtra state populations is composed of mainly South Asian haplogroups with traces of east and w)