Papers reviewed and determined not to be word norm studies. Use the flag icon to report errors or suggest re-inclusion.
18625 papers
This paper describes a new, large scale discourse-level annotation project -- the Penn Discourse TreeBank (PDTB). We present an approach to annotating a level of discourse structure that is based on identifying discourse connectives and their arguments. The PDTB is being built directly on top of the Penn TreeBank and Propbank, thus supporting the extraction of useful syntactic and semantic features and providing a richer substrate for the development and evaluation of practical algorithms.
The paper presents an unlexicalized probabilistic parsing model for German trained on the Negra treebank. Evaluation is performed with respect to constituency and dependency measures. It is observed that existing models based on Parent Encoding and Markovization optimize for constituency measures at the expense of dependency performance (at least in German). Several linguistically inspired transformation and annotation schemes are proposed which do help with dependency measures. Finally, it is shown that performance compares well with published results for German.
When complete, NomBank will provide annotation of noun arguments in Penn Treebank II (PTB). In PropBank, University of Pennsylvania annotators provide similar information for verbs. Given nominalization/verb mappings, the combination of NomBank and PropBank allows for generalization of arguments across parts of speech. This paper describes our annotation task including factors which make assigning role labels to noun arguments a challenging task. 1.
Criterion concept maps developed by instructors in the C-TOOLS project contain numerous expert-generated or “correct” propositions manually created by expert users connecting two object or subject phrases together with a linking phrase. WordNet, an electronic lexical database, was then used to construct additional proposition derivatives by supplying different linking phrases in place of those originally specified by users. Derived propositions were made from original propositions by substituting linking phrase verbs with troponyms, synonyms, and antonyms. During the past year some 1298 students created concept maps (with 35404 propositions) aided by automatic grading and the WordNet® propositions. We have now studied how successful WordNet was at creating valid additional linking words like those generated by experts. By comparing manual assessments of derived propositions to manual assessments of original propositions, the persistence of correctness was evaluated for determinative factors such as frequencies and senses of usage. Results from data analyses are compared to parameters for WordNet usage in order to better determine the potential for the refining of concept map grading algorithms. An empirical approach to word sense relationships is presented as an iterating step to progressively prototype and test optimizations.
This paper introduces the Prague Czech-English Dependency Treebank (PCEDT), a new Czech-English parallel resource suitable for experiments in structural machine translation. We describe the process of building the core parts of the resources – a bilingual syntactically annotated corpus and translation dictionaries. A part of the Penn Treebank has been translated into Czech, the dependency annotation of the Czech translation has been done automatically from plain text. The annotation of Penn Treebank has been tranformed into dependency annotation scheme. A subset of corresponding Czech and English sentences has been annotated by humans. First experiments in Czech-English machine translation using these data have already been carried out. The resources being created at Charles University in Prague are scheduled for release as Linguistic Data Consortium data collection in 2004. 1.
We present the first application of the head-driven statistical parsing model of Collins (1999) as a simultaneous language model and parser for large-vocabulary speech recognition. The model is adapted to an online left to right chart-parser for word lattices, integrating acoustic, n-gram, and parser probabilities. The parser uses structural and lexical dependencies not considered by n-gram models, conditioning recognition on more linguistically-grounded relationships. Experiments on the Wall Street Journal treebank and lattice corpora show word error rates competitive with the standard n-gram language model while extracting additional structural information useful for speech understanding.
We describe an approach to two areas of biomedical information extraction, drug development and cancer genomics. We have developed a framework which includes corpus annotation integrated at multiple levels: a Treebank containing syntactic structure, a Propbank containing predicate-argument structure, and annotation of entities and relations among the entities. Crucial to this approach is the proper characterization of entities as relation components, which allows the integration of the entity annotation with the syntactic structure while retaining the capacity to annotate and extract more complex events. We are training statistical taggers using this annotation for such extraction as well as using them for improving the annotation process.
In natural language processing a huge amount of structured data is constantly used for the extraction and presentation of grammatical structures in sentences. For example the Chinese Treebank corpus developed at the Institute of Information Science Academia Sinica Taiwan is a semantically annotated corpus that has been used to help parse and study Chinese sentences. In this setting users usually use structured tree patterns instead of keywords to query the corpus.
The Penn Discourse TreeBank (PDTB) is a new resource built on top of the Penn Wall Street Journal corpus, in which discourse connectives are annotated along with their arguments. Its use of standoff annotation allows integration with a stand-off version of the Penn TreeBank (syntactic structure) and PropBank (verbs and their arguments), which adds value for both linguistic discovery and discourse modeling. Here we describe the PDTB and some experiments in linguistic discovery based on the PDTB alone, as well as on the linked PTB and PDTB corpora.
While alignment of texts on the sentential level is often seen as being too coarse, and word alignment as being too fine-grained, bi- or multilingual texts which are aligned on a level inbetween are a useful resource for many purposes. Starting from a number of examples of non-literal translations, which tend to make alignment difficult, we describe an alignment model which copes with these cases by explicitly coding them. The model is based on predicateargument structures and thus covers the middle ground between sentence and word alignment. The model is currently used in a recently initiated project of a parallel English-German treebank (FuSe), which can in principle be extended with additional languages.
In this paper we present a proposal to extend WordNet-like lexical databases by adding information about the co-occurrence of word meanings in texts. More specifically we propose to add phrasets, i.e. sets of free combinations of words which are recurrently used to express a concept (let's call them Recurrent Free Phrases). Phrasets are a useful source of information for different NLP tasks, and particularly in a multilingual environment to manage lexical gaps. At least a part of recurrent free phrases can also be represented through a new set of syntagmantic (lexical and semantic) WordNet relations.
Despite much research, the distinctive personality characteristics of entrepreneurs are yet to be established and the influence of personality on entrepreneurial behaviour remains unclear. This is particularly evident in our understanding of the personal response of entrepreneurs to business failure. In this thesis the Life Story Model of Identity proposed by McAdams' (1993; McAdams & Pals, 2006) narrative theory of personality formed the main theoretical approach to investigating these two related aspects in the psychological understanding of entrepreneurs. This model overcomes some of the limitations of previous personality research by permitting investigation of personality within the entrepreneurial environment and provides a wholistic and complex view of personality as expressed in the entrepreneurs' own words. The model's qualitative methodology and theoretical emphasis on personal meaning making also rendered it most suitable for exploring entrepreneurs' personal response to business failure. McAdams' (1993; McAdams & Pals, 2006) Life Story Interview was employed to explore the self-narrative identities of 40 highly successful entrepreneurs (39 males, one female). Participants were managing directors of businesses sourced from two lists of the fastest growing small to medium companies in Australia, as compiled by the Australian business magazine, the 'Business Review Weekly'. Participants were the founders of their businesses, and had been pursuing entrepreneurship for at least five years. A broad range of business sectors were represented, including computer services, manufacturing, engineering and communications. Prior to interviews, participants completed the Life Story Interview Questionnaire (LSIQ), which was an adapted form of the Life Story Interview that requested written responses to open-ended questions about the content of participants' life stories. A section requesting affective ratings for key events, derived from Herman’s (Hermans & Hermans-Jansen, 1995) Extended List of Affect Terms, was included to further the exploration of life story themes. A second questionnaire, comprised of measures of personality and a measure of psychological symptoms was also completed. During interviews, participants' responses to the LSIQ were discussed, concentrating on further investigation of the key events that defined their life stories. Findings revealed a prototypical life story of the entrepreneur, highlighting distinctive, commonly shared personality characteristics, with much of their selfnarrative identity grounded in experiences within the entrepreneurial environment. The prototypical life story contained a core theme with an agentic-type emphasis on strengthening the self, and a lesser theme with a communion-type emphasis on valuing relationships. Each of these themes comprised two further themes. The selfstrengthening theme included a redemptive theme of overcoming difficulties in a way that left the protagonist feeling stronger and more able to influence their environment, and a positively toned theme of drawing strength and confidence in one's abilities from achievements and successes. The relational theme included a redemptive theme of responding to private relationship difficulties and losses in one area by strengthening other private relationships, and a negatively toned, sometimes contaminated theme, of experiencing either private or professional relational difficulties and losses as irresolvable. The resulting prototypical life story of the entrepreneur was a story centred upon overcoming adversity and celebrating personal achievement, of confirming and boosting confidence in one’s abilities and a sense of personal power to influence their environment. Running parallel to this main storyline was a less prominent plot involving the importance of relationships, with difficulties and losses sometimes redeemed and sometimes left unresolved. To investigate the impact of business failure, participants were asked to describe their experience of business failure as a key life story event during the Life Story Interview. Additional open-ended questions explored important elements of their critical and retrospective responses. A commonly shared personal response was evidenced. Despite being strongly identified with their business at the time, the failure was evaluated in business rather than personal terms. The causes were most often attributed to a combination of internal and external factors, but business recovery was attributed exclusively to their own actions. The self-narrative meanings given to this event centred upon overcoming the business failure in a self-strengthening way as either: mastering business conflict situations; learning entrepreneurial skills; or affirming entrepreneurial self-confidence. In the midst of the failure, most entrepreneurs remained highly optimistic about their chances of success in the future, based largely upon confidence in their ability to bring about business recovery. Practised coping skills were used to manage negative feelings arising from the failure, and an active problem-solving approach was adopted. When reflecting upon their experience, there was an absence of rumination and regret about the business failure. Instead, it was regarded as an inevitable and even welcome event that provided valuable entrepreneurial learning. In making sense of the failure in relation to the rest of their self-narrative identity, most entrepreneurs were able to integrate its meaning within their larger self-story. This was done by relating the business’ recovery to a story of overcoming obstacles, or by containing the business’ failure within a story of either repeated success or sustained self-confidence in one's ability. It was concluded that these entrepreneurs shared a particular type of selfnarrative identity that was conducive to the pursuit of entrepreneurship; positively influencing their behaviour within the entrepreneurial environment and having particular relevance to how they personally responded to business failure. These findings advance understanding of the personality of entrepreneurs, and begin to inform what constitutes a constructive personal response to the event of business failure.
Tree-based approaches to alignment model translation as a sequence of probabilistic operations transforming the syntactic parse tree of a sentence in one language into that of the other. The trees may be learned directly from parallel corpora (Wu, 1997), or provided by a parser trained on hand-annotated treebanks (Yamada and Knight, 2001). In this paper, we compare these approaches on Chinese-English and French-English datasets, and find that automatically derived trees result in better agreement with human-annotated word-level alignments for unseen test data.
We present a system for unsupervised tagging of words into classes produced by a distributional clustering technique called co-clustering. A hidden Markov model (HMM), trained on the high-frequency terms in the lexicon, is used to "tag" occurrences of low-frequency terms. In experiments using the Wall Street Journal portion of the Penn Treebank, we show that previously reported problems in using Baum-Welch estimation for part-of-speech tagging do not occur in this context. We also show how state-level term emission models can be augmented to account for morphological patterns using features automatically derived from the output of co-clustering. Finally, we consider an alternative means of extending the coverage of the lexicon, in which low-frequency terms are added to the lexicon as types, and compare this approach with the token-level assignments made by the HMM.
There are three types of sentences that form all existing natural languages: verbal sentences (e.g."I read the book."),copulative sentences (e.g."The book is on the table."),and existential sentences (e.g."There is a book on the table.").Syntactic and semantic recognition of these sentence types are crucially important in computational linguistics although there has not been any significant work towards this end.This thesis, in an attempt to fill this evident gap, is on identifying and assigning semantic categories of Turkish existential sentences in print.Existential sentences in Turkish are minimally characterized by the two existential particles var, meaning there is/are, and yok, meaning there is/are no.In addition to these most basic meanings, other senses of existential particles are possible, which can be categorized into groups such as case existentials and possession existentials.Our system does shallow semantic parsing in defining the predicate-argument relationships in an existential sentence on a word-byword basis, via utilizing Support Vector Machines, after which it proceeds with the semantic categorization of the whole sentence.For both of these tasks, our system produces promising results, in terms of accuracy and precision/recall, respectively.Part of this research contributes to the annotation of the METU-Sabanc Turkish Treebank with semantic information.
Alcohol has been shown to have both impairing and facilitating effects on memory, depending on the sequencing of learning and ingestion of the drug. Its effects on memory for emotional material, however, have not been shown reliably. The current experiment sought to investigate the effects of alcohol on later recall of emotional and neutral events experienced before and after alcohol drinking. Using an incidental-learning paradigm, alcohol (0.65 g/kg) or placebo was administered in a double-blind randomized design to 34 participants, between two learning phases in which they viewed and rated positive, negative and neutral images. The drug's effects on memory were assessed in a surprise test of free recall. In addition, impact of alcohol on ratings of mood states, and of valence and arousal that the pictures evoked, was examined. Alcohol facilitated memory for material seen before, and impaired memory for material seen after, its administration. Furthermore, under alcohol, emotional images in the first set were more recalled over neutral than in the second set, indicating a higher retrograde facilitation for emotional than for neutral material. Alcohol improved positive mood states but had no effect on negative mood states. Evaluation of pictures with regard to valence showed an increase of the ratings for the positive and neutral images after alcohol and a decrease after placebo. No drug effects were found for arousal ratings. Whether a picture was likely to be remembered or not (tested only for set 2) was dependent on the intensity of the arousal but not of the valence that the picture evoked in the participants. Pictures that were rated high in arousal were also remembered better, and this effect was irrespective of alcohol or placebo ingestion. These data have shown that alcohol elicits retrograde facilitation and anterograde impairment for emotional materials. Furthermore, these data demonstrate that alcohol selectively facilitates memories for emotional events experienced before administration, and suggest a possible explanation for the reinforcing effects of drinking.
Most statistical parsers have used the grammar induction approach, in which a stochastic grammar is induced from a treebank. An alternative approach is to induce a controller for a given parsing automaton. Such controllers may be stochastic; here, we focus on greedy controllers, which result in deterministic parsers. We use decision trees to learn the controllers. The resulting parsers are surprisingly accurate and robust, considering their speed and simplicity. They are almost as fast as current part-ofspeech taggers, and considerably more accurate than a basic unlexicalized PCFG parser. We also describe Markov parsing models, a general framework for parser modeling and control, of which the parsers reported here are a special case.
When two targets are presented in rapid serial visual presentation (RSVP) streams, report accuracy for the second target is low when it is presented less than 500 ms after the first target (an attentional blink). In contrast, report accuracy of single targets presented in RSVP streams is high, given that RSVP distractors do not typically deplete the attentional resources needed to perform the target task. In our experiment participants were asked to report the identity of a single color word target that appeared in each RSVP stream (a 10 AFC response). A pseudo-target word that appeared to be a normal distractor was presented 1, 2, 3, 4, 5, or 8 items before the target word. This pseudo-target was of neutral valence (binder, aisle), negative valence but low arousal (boredom, sad), positive valence (cheer, glad), or high arousal and variable valance (orgasm, shit). The arousing/taboo words led to lower overall accuracy on the target task, and set-off an AB to a subsequent color word target, despite the fact that they appeared as distractors. However, no AB was evidenced with negative, positive, or neutral word distractors. Target accuracy was predicted by the arousal ratings, but not the valence ratings (positive-negative) that participants gave to the words. The presence of this emotion blindness provides evidence that arousing words grab attention in a bottom-up manner, sometimes at the expense of top-down target search.
The PropBank project is creating a corpus of text annotated with information about basic semantic propositions. PropBank I (Kingsbury & Palmer, 2002) added a layer of predicateargument information, or semantic roles, to the syntactic structures of the English Penn Treebank. This paper presents an overview of the second phase of PropBank Annotation, PropBank II, which is being applied to English and Chinese, and includes (Neodavidsonian) eventuality variables, nominal references, sense tagging, and connections to the Penn Discourse Treebank (PDTB), a project for annotating discourse connectives and their arguments. 1
This paper presents a novel algorithm for computing similarity between very short texts of sentence length. It will introduce a method that takes account of not only semantic information but also word order information implied in the sentences. Firstly, semantic similarity between two sentences is derived from information from a structured lexical database and from corpus statistics. Secondly, word order similarity is computed from the position of word appearance in the sentence. Finally, sentence similarity is computed as a combination of semantic similarity and word order similarity. The proposed algorithm is applied to a real world domain of conversational agents. Experimental results demonstrated that the proposed algorithm reduces the scripter's effort to devise rule base for conversational agent.
Abstract. This paper describes experiments on using inductive machine learning to guide a deterinistic dependency parser for unrestricted natural language text. Using data from a small treebank of Swedish, an eager probabilistic learning algorithm is used to induce context-sensitive parse tables. Evaluation shows a significant improvement over the baseline, which uses a table without contextual information. 1
OBJECTIVE: Transcutaneous electrical nerve stimulation (TENS) is a technique widely used in clinical practice to control pain, although its clinical efficacy remains controversial. Though many mechanisms have been proposed for its analgesic effects, there is a conspicuous lack of experimentally controlled research investigating whether TENS analgesia is related to its effects on the sympathetic nervous system (SNS). METHODS: Using an established psychophysiological paradigm, the present study investigated the effects of high-frequency/low-intensity TENS, low-frequency/high-intensity TENS, and sham TENS on the perception of experimental pain and SNS function in healthy volunteers. Measures of heart rate, digital pulse volume, and skin conductance were recorded during a 20-minute TENS stimulation period and in anticipation of a series of painful electric shocks prior to and following TENS stimulation. Healthy volunteers rated the intensity of the shocks using a 0-10-point verbal pain rating scale. RESULTS: The three TENS conditions failed to differentially effect SNS responses during either the 20-minute TENS treatment period or the shock anticipation periods, and TENS did not affect ratings of pain intensity to the shock stimuli. CONCLUSIONS: While these results may not generalize to acute or chronic pain patients, within the limitations of the present experimental paradigm, no support was found for TENS affecting either SNS function or acute experimental pain perception.
INTRODUCTION Several authors described cases of dissociated impairment in naming nouns and verbs. There are four accounts of this dissociation: (i) patients may have purely lexical damage, which selectively affects verbs or nouns at a late stage of the linguistic processing (phonological or orthographic lexicons) (Rapp & Caramazza, 2002); (ii) the damage affects a lexical device, either at an ortographic-phonological modality-specific level (the lexeme; Levelt et al., 1999) or at a unitary lexical-syntactic level (the lemma) (Berndt et al., 1997); (iii) N-V dissociation arises from a semantic damage (Bird, Howard & Franklin, 2000); (iv) N-V dissociation is due to syntactic damage (Friedmann & Grodzinsky, 2000). To disentangle imageability and grammatical class effects, a new task was developed allowing to elicit nouns and verbs with identical imageability ratings in a sentence context. The results obtained will permit to address the following three questions: Does imageability play a role in determining N-V dissociation? If so, is imageability the unique cause of dissociation? If there is additional damage, at which level of linguistic processing does it take place? METHODS Twelve Italian aphasic patients and 11 normal controls participated in the study. Nouns and Verbs Retrieval in a Sentence Context (NVR-SC). Forty-five pairs of sentences denoting the same event, either using a noun or the corresponding verb (e.g. the evasion/to evade) were used. The first sentence was presented in complete form, while a gap was left in the second sentence, to be completed with the target word. For each pair of sentences two different conditions were employed, one triggering a verb and one triggering a noun. E.g. V-N: The prisoner was dreaming to evade The prisoner was dreaming the......... N-V: The prisoner was dreaming the evasion The prisoner was dreaming to......... The performance in the NVR-SC task was compared to that obtained on a classic picture naming task (50 nouns and 50 verbs). Statistical methods. Logistic regression analysis (LRA) was applied to the profiles of the patients, making it possible to study the effects of the lexical-semantic variables in univariate and multivariate linear models. RESULTS Picture naming task All patients with predominant verb deficit also have an imageability effect. In eight patients the grammatical class effect was no longer significant after introducing imageability in the statistical design (bivariate LRA). NVR-SC task Two of the verb-impaired patients in the picture naming task maintained the predominant verb deficit also in the NVR-SC task. In eight patients, the difference between nouns and verbs was no longer significant. In two patients, a paradoxical dissociation (V>N) emerged. Group analysis: performance on nouns and verbs across naming tasks (Figure 1) Patients named actions in the NVR-SC task better than in the picture naming task (58% correct versus 37%; p<.001). On the contrary, the naming of objects in the picture naming task was better than in the NVR-SC task (76% versus 61%; p<.05). DISCUSSION As expected, imageability effect is highly associated with noun-superiority. This result may have two explanations. (a) Since nouns are generally more imaginable than verbs, the imageability effect may cause noun-superiority. But, imageability alone cannot completely account for predominant verb impairment. In fact, four patients have predominant verb impairment even when imageability has been partialled out (bivariate LRA) and two patients are still dissociated in the NVR-SC task (where imageability of nouns and verbs was perfectly matched). Therefore, additional damage must be hypothesized. This would account for the fact that nouns are named better in the picture naming task, and verbs in the NVR-SC task. The former result is explained by the reduced imageability ratings; the latter may only be explained by localizing the additional damage at a central, lexical-syntactic level (i.e. the lemma). This explanation would account for the better performance on verbs in the NVR-SC task: the sentence frame and the correspondent noun may provide the patients with those information lost with the lemma damage, i.e. number of arguments and thematic roles. Nonetheless, this may account also for the fact that noun retrieval is not enhanced in the NVR-SC task: the thematic grid is not useful to retrieve nouns, since it is not a crucial aspect of noun lexical representation. In addition, aphasic patients may use a compensatory strategy to face their deficit. Since the thematic grid may be inferred from a mental image, this strategy will probably rely on visual representations of actions. Thus, the effectiveness of the compensatory strategy increases in relation to imageability (Luzzatti & Chierchia, 2002). (b) When left hemisphere language areas are completely damaged, lexical representations located in the right hemisphere emerge. This emergence is limited to high-frequency concrete nouns (Coltheart, 2000).
This paper presents the current status of the French treebank developed at Paris 7 (Abeillé et al., 2003a). The corpus comprises 1 million words from the newspaper le Monde, fully annotated and disambiguated for parts of speech, inflectional morphology, compounds and lemmas, and syntactic constituents. It is representative of contemporary normalized written French, and covers a variety of authors and subjects (economy, literature, politics, etc.), with extracts from newspapers ranging from 1989 to 1993. It has been used by computational linguists to train and evaluate taggers, parsers and lemmatizers, as well as by psycholinguists to extract lexical and syntactic preferences (Pynte et al., 2001). It is now being enriched with functional information, and used for parsing evaluation. 1. The French treebank Similarly to the Penn TreeBank, we have annotated both parts of speech and constituents. Differently from the Penn Treebank, we have also annotated compounds, lemmas and inflectional morphology. Our annotation choices are meant to be linguistically motivated and compatible with various linguistic theories. We have chosen surface-based annotations, with no empty categories (Abeillé and Clément,
This paper discusses an annotation scheme for Korean null pronouns, which were used in annotating three kinds of Korean text corpora including Penn Korean Treebank. In annotating the corpora, null pronouns and their antecedents were marked up for their type and reference, with coreference relation tracked by numeric identifiers. Based on the annotation scheme, an outline of a potential pronoun resolution strategy is also proposed. The resulting dataset of annotated text is rather small at 11,834 words; we hope the null pronoun classification and annotation scheme proposed in this study will serve as a basis in developing a large-scale annotated corpus in the future.
This paper presents the preliminary analysis of Kannada WordNet and the set of relevant computational tools. Although the design has been inspired by the famous English WordNet, and to certain extent, by the Hindi WordNet, the unique features of Kannada WordNet are graded antonyms and meronymy relationships, nominal as well as verbal compoundings, complex verb constructions and efficient underlying database design (designed to handle storage and display of Kannada unicode characters). Kannada WordNet would not only add to the sparse collection of machine-readable Kannada dictionaries, but also will give new insights into the Kannada vocabulary. It provides sufficient interface for applications involved in Kannada machine translation, spell checker and semantic analyser.
In the present paper detailed schemes are proposed for an annotated corpus and a lexical database of Modern Hebrew. They are meant to be primary linguistic sources for more empirical studies of the grammatical and lexical structure of Modern Hebrew to shed light on aspects and phenomena hitherto unknown. XML (Extensible Markup Language) is chosen as their storage format because of its machine- and humanreadability, crossplatform-compatibility, crosslinguistic-compatibility, self-descriptiveness and capability of nesting structure. The corpus will be annotated in four levels, i.e., syntactically, morphosyntactically, lexically and morphologically. The lexical database will include modules of morphosyntax, inflection, word-formation and syntacticosemantics. The data will be recorded in Unicode, whether in Hebrew characters or in Latin transcription. Although countless numbers of revisions have been made since the idea of building these two sources in XML was first born a few years ago, what is proposed here is essentially by one individual. It might, therefore, need minor (or even major) revisions and/or expansions.
A part-of-speech (POS) tagged corpus was built on research abstracts in biomedical domain with the Penn Treebank scheme. As consistent annotation was difficult without domain-specific knowledge we made use of the existing term annotation of the GENIA corpus. A list of frequent terms annotated in the GENIA corpus was compiled and the POS of each constituent of those terms were determined with assistance from domain specialists. The POS of the terms in the list are pre-assigned, then a tagger assigns POS to remaining words preserving the pre-assigned POS, whose results are corrected by human annotators. We also modified the PTB scheme slightly. An inter-annotator agreement tested on new 50 abstracts was 98.5%. A POS tagger trained with the annotated abstracts was tested against a gold-standard set made from the interannotator agreement. The untrained tagger had the accuracy of 83.0%. Trained with 2000 annotated abstracts the accuracy rose to 98.2%. The 2000 annotated abstracts are publicly available. 1.
Systemic functional linguistics offers a grammar that is semantically organised, so that salient grammatical choices are made explicit. This paper describes the explication of these choices through the conversion of the Penn Treebank into a systemic functional grammar corpus. Developing such a resource can help connect work in natural language processing to a significant body of research dealing explicitly with the issue of how lexical and grammatical selections create meaning.
Rating agencies' track record is good in developed countries but poor in emerging economies. Why? Given the almost-monopolistic structure of the industry, we conjecture that agencies might underinvest in information gathering. We propose an indicator quantifying the agencies' effort to gather information and assess whether greater effort affects rating levels. We detect: (i) absolute underinvestment for non-OECD sovereigns (less effort in spite of greater opaqueness); (ii) relative underinvestment for non-OECD firms compared with OECD ones (though the former receive a larger effort, more intense effort boosts firm ratings in non-OECD countries while depressing them in OECD countries).
How visual cues reveal animacy and social meaning to moving displays is an open question. We investigated this question using techniques of computer vision and image processing that enable full video displays of human action to be decomposed and subsequently transformed into abstract representations of the movement. Using the multimedia analysis program Eyesweb (Infomus lab, U. Genoa), we created visual displays of the same interaction between two dancers performing a modern dance. Four conditions were created that reduced the visual information available. These were: (1) full video recording; (2) body silhouettes; (3) motion of the center of each dancer represented by a single small block changing size in accordance with the motion energy of the dancer; (4) motion of the center of the dancer represented by a single small block that did not change size. Using a between subjects design, participants viewed a two-minute display of a single condition. While viewing, they were asked to provide a continuous measurement of arousal by moving a slider up and down. After this they were asked to write a description of what they had seen. Average arousal ratings were obtained for each condition and analysis revealed higher average ratings of arousal for the full video and silhouette conditions. The free response data was analyzed for spontaneous mention of terms relating to animacy of the display elements. These results indicated that animacy terms were used for the full video and silhouette conditions but not the block conditions. The results are discussed in relation to previous methods used to create stimuli resulting in the attribution of animacy and social meaning to abstract displays of motion.
Information exchange is increasing rapidly with the advent of globalization. As the language spoken by the most people in today's world, Chinese will play an important role in information exchange in the future. Therefore, we need an efficient and practical means to access the increasingly large volume of Chinese data. This thesis describes a target-dominant Chinese-English machine translation system, which can translate a given Chinese news sentence into English. We conjecture that we can improve the state of the art of MT using a TDMT approach. This system has participated in the NIST (National Institute of Standards and Technology) 2004 machine translation competition. Experimental results on Penn Chinese Treebank corpus show that a machine translation system adopting a target-dominant approach is promising.
Several studies have shown that facial attractiveness is positively correlated with both familiarity and typicality. Here we manipulated the familiarity of typical and distinctive faces to measure the effect on attractiveness. In our first experiment, we collected ratings of attractiveness, distinctiveness, and familiarity using three different groups of participants. Our stimuli included 84 images of female faces, presented in a full-face view. We replicated the finding that attractiveness ratings negatively correlate with distinctiveness ratings. In addition, we showed that attractiveness ratings were positively correlated with familiarity ratings. In our second experiment, we demonstrated that increasing exposure to faces increases their attractiveness, although there was no differential effect of exposure on typical and distinctive faces. Our results suggest that episodic familiarity affects attractiveness ratings independently of general or structural familiarity. The implications of our findings for the 'face-space' model are discussed.
An efficient bit-vector-based CKY-style parser for context-free parsing is presented. The parser computes a compact parse forest representation of the complete set of possible analyses for large treebank grammars and long input sentences. The parser uses bit-vector operations to parallelise the basic parsing operations. The parser is particularly useful when all analyses are needed rather than just the most probable one.
In this paper we introduce a naive algorithm for nondeterminisctic LTAG derivation tree extrac-tion from the Penn Treebank and the Proposi-tion Bank. This algorithm is used in the EM models of LTAG Treebank Induction reported in (Shen and Joshi, 2004). Given the trees in the Penn Treebank with PropBank tags, this algorithm generates shared structures that al-low efficient dynamic programming in the EM models. 1
We investigated subjective and hemodynamic responses towards disgust-inducing, fear-inducing, and neutral pictures in a functional magnetic resonance imaging study. Within an interval of 1 week, 24 male subjects underwent the same block design twice in order to analyze possible response changes to the repeated picture presentation. The results showed that disgust-inducing and fear-inducing scenes provoked a similar activation pattern in comparison to neutral scenes. This included the thalamus, primary and secondary visual fields, the amygdala, the hippocampus, and various regions of the prefrontal cortex. During the retest, the affective ratings hardly changed. In contrast, most of the previously observed brain activations disappeared, with the exception of the temporo-occipital activation. An additional analysis, which compared the emotion-related activation patterns during the two presentations, showed that the responses to the fear-inducing pictures were more stable than the responses to the disgust-inducing ones.
Abstract. WordNet is an electronic lexical database structured around psychological and linguistic principles. As such, it should play a part in any e ort to integrate cognitive factors into knowledge based systems. Yet, some of its basic assumptions have been attacked, and the suggestion made that a major restructuring would make it more cognitively transparent. We investigate these allegations from a psycholinguistic perspective and conclude that WordNet is in fact rigorous in terms of the cognitive principles it embodies. What is lacking is a methodology for translating the explicit and implicit knowledge in WordNet into a usable, formal ontologies. We show some ways in which WordNet should be extended to facilitate this process. We agree that WordNet is not in itself ready for use as a formal ontology, but we argue that it is an invaluable tool for describing the conceptualized structure of our world, and should be used as a fundamental resource. 1
Although previous functional brain imaging studies have found that the insula responds selectively to facial expressions of disgust, it remains unclear whether the insula responds selectively to disgust-inducing pictures. In this fMRI study, healthy volunteers viewed pictures of contamination, human mutilation, attacks and neutral scenes during scanning, and then rated pictures for the 'basic' emotions. The anterior insula responded to contamination and mutilation but not attacks, while the ventral visual areas responded to attacks and mutilations more strongly than contamination. The above activations were predicted by disgust and arousal ratings respectively. Additionally, mutilations uniquely activated the right superior parietal cortex. These results support selective disgust processing at the insula, and suggest distinct neural responses to contamination and mutilation.
The aim of this study is to highlight what kind of information distinguishes abstract and concrete conceptual knowledge in different aged children. A familiarity-rating task has shown that 8-year-olds judged concrete concepts as very familiar while abstract concepts were judged as much less familiar with ratings increasing substantially from age 10 to age 12, according to literature showing that abstract terms are not mastered until adolescence (Schwanenflugel, 1991). The types of relation elicited by abstract and concrete concepts during development were investigated in an association production task. At all considered age levels, concrete concepts mainly activated attributive and thematic relations as well as, to a much lesser extent, taxonomic relations and stereotypes. Abstract concepts, instead, elicited mainly thematic relations and, to a much lesser extent, examples and taxonomic relations. The patterns of relations elicited were already differentiated by age 8, becoming more specific in abstract concepts with age.
A critical issue in the study of speech motor control is the identification of the mechanisms that generate the temporal flow of serially ordered articulatory events. Two staged models of serial ordered events (Lashley, 1951; Lindblom, 1963) claim that time controls events whereas dynamic models predict a relative relation between time and space. Each of these models predicts a different relation between the acoustic measures of formant frequency and segmental duration. The most recent method described herein provides a sensitive index of speech deterioration which is both acoustically robust and phonetically systematic. Both acoustic and magnetic resonance imaging measures were used to describe the speech disturbance in two neurologically distinct groups of cerebellar ataxia: Friedreich’s ataxia and olivo-ponto cerebellar ataxia. The speaking task was designed to elicit six different prosodic conditions and four prosodic contrasts. All subjects read the same syllable embedded in a sentence, under six different prosodic conditions. Pair-wise comparisons derived from the six conditions were used to describe (1) final lengthening, (2) phrasal accent, (3) nuclear accent and (4) syllable reduction. An estimate of speech deterioration as determined by individual and normal subects’ acoustic values of syllable duration, formant and fundamental frequencies was used in correlation analyses with magnetic resonance imaging ratings.