Papers reviewed and determined not to be word norm studies. Use the flag icon to report errors or suggest re-inclusion.
16504 papers
Current automatic wrappers using DOM tree and visual properties of data records to extract the required information from the deep web generally have limitations such as the inability to check the similarity of tree structures accurately. Our study shows that data records located in the deep web do not only share similar visual properties and tree structures, but they are also related semantically in their contents. As such we are able to propose an ontological technique using existing lexical database for English (WordNet) for the extraction of data records from deep web pages. Wrappers designed based on ontological technique are able to reduce the number of potential data regions identified for data extraction, thus improve the data extraction accuracy. In this study, we use visual cue from the underlying browser rendering engine to locate and extract the relevant data region from the deep web by measuring the text and image sizes of data records. Experimental results show that our technique is robust and performs better than the existing state of the art wrappers. Unlike existing ontological based wrappers, our wrapper is domain independent and is able to extract wide range of data records with different structures.
We present the first evaluation of the utility of automatic evaluation metrics on surface realizations of Penn Treebank data. Using outputs of the OpenCCG and XLE realizers, along with ranked WordNet synonym substitutions, we collected a corpus of generated surface realizations. These outputs were then rated and post-edited by human annotators. We evaluated the realizations using seven automatic metrics, and analyzed correlations obtained between the human judgments and the automatic scores. In contrast to previous NLG meta-evaluations, we find that several of the metrics correlate moderately well with human judgments of both adequacy and fluency, with the TER family performing best overall. We also find that all of the metrics correctly predict more than half of the significant systemlevel differences, though none are correct in all cases. We conclude with a discussion of the implications for the utility of such metrics in evaluating generation in the presence of variation. A further result of our research is a corpus of post-edited realizations, which will be made available to the research community. 1
This paper proposes a dependency parsing method that uses bilingual constraints to improve the accuracy of parsing bilingual texts (bitexts). In our method, a targetside tree fragment that corresponds to a source-side tree fragment is identified via word alignment and mapping rules that are automatically learned. Then it is verified by checking the subtree list that is collected from large scale automatically parsed data on the target side. Our method, thus, requires gold standard trees only on the source side of a bilingual corpus in the training phase, unlike the joint parsing model, which requires gold standard trees on the both sides. Compared to the reordering constraint model, which requires the same training data as ours, our method achieved higher accuracy because of richer bilingual constraints. Experiments on the translated portion of the Chinese Treebank show that our system outperforms monolingual parsers by 2.93 points for Chinese and 1.64 points for English. 1
RATIONALE: Acute tryptophan depletion (ATD) decreases levels of central serotonin. ATD thus enables the cognitive effects of serotonin to be studied, with implications for the understanding of psychiatric conditions, including depression. OBJECTIVE: To determine the role of serotonin in conscious (explicit) and unconscious/incidental processing of emotional information. MATERIALS AND METHODS: A randomized, double-blind, cross-over design was used with 15 healthy female participants. Subjective mood was recorded at baseline and after 4 h, when participants performed an explicit emotional face processing task, and a task eliciting unconscious processing of emotionally aversive and neutral images presented subliminally using backward masking. RESULTS: ATD was associated with a robust reduction in plasma tryptophan at 4 h but had no effect on mood or autonomic physiology. ATD was associated with significantly lower attractiveness ratings for happy faces and attenuation of intensity/arousal ratings of angry faces. ATD also reduced overall reaction times on the unconscious perception task, but there was no interaction with emotional content of masked stimuli. ATD did not affect breakthrough perception (accuracy in identification) of masked images. CONCLUSIONS: ATD attenuates the attractiveness of positive faces and the negative intensity of threatening faces, suggesting that serotonin contributes specifically to the appraisal of the social salience of both positive and negative salient social emotional cues. We found no evidence that serotonin affects unconscious processing of negative emotional stimuli. These novel findings implicate serotonin in conscious aspects of active social and behavioural engagement and extend knowledge regarding the effects of ATD on emotional perception.
Tree-based translation models, which exploit the linguistic syntax of source language, usually separate decoding into two steps: parsing and translation. Although this separation makes tree-based decoding simple and efficient, its translation performance is usually limited by the number of parse trees offered by parser. Alternatively, we propose to parse and translate jointly by casting tree-based translation as parsing. Given a source-language sentence, our joint decoder produces a parse tree on the source side and a translation on the target side simultaneously. By combining translation and parsing models in a discriminative framework, our approach significantly outperforms a forestbased tree-to-string system by 1.1 absolute BLEU points on the NIST 2005 Chinese-English test set. As a parser, our joint decoder achieves an F1 score of 80.6 % on the Penn Chinese Treebank. 1
This paper presents an efficient approach to use small syntactic category for constructing large tree with the assist of linguistic knowledge for parsing Thai language. VB-EM algorithm is exploited to adjust parameters of any trees for selecting the best tree. Our best result is 70.62% accuracy of bracketing recovery.
Facial electromyography (EMG) is a useful physiological measure for detecting subtle affective changes in real time. A time series of EMG data contains bursts of electrical activity that increase in magnitude when the pertinent facial muscles are activated. Whereas previous methods for detecting EMG activation are often based on deterministic or externally imposed thresholds, we used regime-switching models to probabilistically classify each individual’s time series into latent “regimes” characterized by similar error variance and dynamic patterns. We also allowed the association between EMG signals and self-reported affect ratings to vary between regimes and found that the relationship between these two markers did in fact vary over time. The potential utility of using regime-switching models to detect activation patterns in EMG data and to summarize the temporal characteristics of EMG activities is discussed.
This paper deals with the issue of linguistic database summaries understood as linguistically quantified propositions. We show basic capabilities of Quantirius, our interactive system supporting the mining and the assessment of linguistic summaries in a database. The mining of linguistic summaries is realized by means of the concept of a protoform introduced by Zadeh. The assessment of validity degrees of summaries is done via Zadeh's fuzzy logic based calculus, extended by the use of cardinality patterns and triangular norms. The main part of the paper presents an idea of a further processing of the set of generated summaries. The proposed algorithm is composed of a reduction mechanism of summaries based on linguistic terms inclusion and a reduction of summaries by means of the overlapping unimodal linguistic terms. We then present an idea of the approval threshold computing for the truth degrees of summaries. Finally, we show that summaries generated by different protoforms can be joined in a master-detail-like relation, making an interesting structure of information contained in the database.
Much of the work on Statistical Machine Translation (SMT) from morphologically rich languages has shown that morphological tokenization and orthographic normalization help improve SMT quality because of the sparsity reduction they contribute. In this paper, we study the effect of these processes on SMT when translating into a morphologically rich language, namely Arabic.We explore a space of tokenization schemes and normalization options. We only evaluate on detokenized and orthographically correct (enriched) output. Our results show that the best performing tokenization scheme is that of the Penn Arabic Treebank. Additionally, training on orthographically normalized (reduced) text then jointly enriching and detokenizing the output outperforms training on enriched text.
The Arabic language has a very rich morphology where a word is composed of zero or more prefixes, a stem and zero or more suffixes. This makes Arabic data sparse compared to other languages, such as English, and consequently word segmentation becomes very important for many Natural Language Processing tasks that deal with the Arabic language. We present in this paper two segmentation schemes that are morphological segmentation and Arabic TreeBank segmentation and we show their impact on an important natural language processing task that is mention detection. Experiments on Arabic TreeBank corpus show 98.1 % accuracy on morphological segmentation and 99.4 % on morphological segmentation. We also discuss the importance of segmenting the text; experiments show up to 6F points improvement of the mention detection system performance when morphological segmentation is used instead of not segmenting the text. Obtained results also show up to 3F points improvement is achieved when the appropriate segmentation style is used. 1.
Arabic is a morphologically rich language, which presents a challenge for part of speech tagging. In this paper, we compare two novel methods for POS tagging of Arabic without the use of gold standard word segmentation but with the full POS tagset of the Penn Arabic Treebank. The first approach uses complex tags that describe full words and does not require any word segmentation. The second approach is segmentation-based, using a machine learning segmenter. In this approach, the words are first segmented, then the segments are annotated with POS tags. Because of the word-based approach, we evaluate full word accuracy rather than segment accuracy. Word-based POS tagging yields better results than segment-based tagging (93.93 % vs. 93.41%). Word based tagging also gives the best results on known words, the segmentation-based approach gives better results on unknown words. Combining both methods results in a word accuracy of 94.37%, which is very close to the result obtained by using gold standard segmentation (94.91%). 1.
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), 115-126. \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.
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), 1-3. \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.
French, like most Romance languages, displays both prenominal and postnominal placement of at-tributive adjectives. The fact that choice in position is not random led many linguists (Abeille and Godard (1999); Forsgren (1978); Wilmet (1981)) to propose constraints based on different dimensions of language (syntax, semantics, phonology, morphology, pragmatics...), but most of them are only trends and it is very difficult to draw a general picture of the phenomenon. We thus proposed in a previous work (Anonymous (tted)) a prediction model built on data from the French Treebank corpus (FTB) (Abeille et al. (2003)), along the lines of Bresnan et al. (2007), based on most of the proposed constraints, to test the contribution weight of each of them and the impact of their interaction. Even if results were encouraging, several facts were outlined: first, if introspection leads to the conclusion that most adjectives alternate, usage shows much more fixity: less than 10% of the adjectives in our data are actually in both positions. This suggests that locutors' mental representations for every item are in fact much stronger for a given position compared to its counterpart. Second, some contrainsts dont have a significant effect in our model. Yet, a qualitative exam showed that they are greatly correlated to a position for specific adjectives. For instance, different 'different', which appears equally in both positions, displays for each position a pattern linked to the nature of the determiner, whereas the type of determiner is not relevant at a more general level. More precisely, we observe a strong cooccurrence of a definite determiner with different in anteposition and a similar pattern between the indefinite determiner and the postposed adjective. This indicates that some constraints are relevant, but only for specific adjectives. Like (Bybee and Mcclelland (2005); Goldberg (2006); Croft (2001)) we believe that locutors' knowledge is based on much more specific information on the item, but also on the context in which it appears: formal characteristics of a particular sequence, frequency of use, distributionality... This work focuses on a qualitative study, on another corpus (ER (2010) 147,934,722 tokens), of the adjectives identified as displaying this alternation in the FTB, the aim being to better understand their functioning, and to propose a model that will better handle actual usage in the FTB. Our methodology is inspired by Gries and Stefanowitsch (2004): we search which lexical elements occur in a particular pattern and identify their attraction strength to the construction by means of a statistical analysis. Results show different types of behaviour on a continuum going from very fixed general patterns to more alternation. The first class of adjectives is very close to the fixed position adjectives: they massively prefer a given position, except in a few cases of idiomatic/collocational sequences. For instance, majeur 'major' is always postposed to the noun, unless it is in the sequence majeure partie 'most part'. In a second class, we still see a great preference for a given position but the alternating cases show less fixity. Two patterns appear within this class: the alternate order either corresponds to a use driven by one major constraint, or to a cumulation/interaction of different constraints. For the first pattern, the constraint involved is not necessarilly the same for every adjective. It can be semantically grounded, which usually leads to distinct nominal paradigms combining with the adjective given its position (e.g. ancien: 'ancien+N' means 'former', 'N+ancien' means 'old'), or it can be based on other devices, as illustrated by different. The data of the ER corpus differs from the FTB by the fact that it shows a strong preference for anteposition. The findings concerning the definite/indefinite nature of the NP were however confirmed, with 97,5% cases of definite in anteposition and 96% of indefinite in postposition. The adjective nouveau 'new' illustrates the second pattern: it prefers anteposition, but postposition is favoured when combined with a concrete noun, or ehanced when the NP is complement of a preposition. The third class concerns adjectives for which the general pattern shows a much weaker preference, if any, for one position over the other. There are however differences in usage for each position. The two patterns outlined for the preceeding class may also apply here: for instance, the placement of principal 'main' depends on a cumulation of information based on different grounds, whereas the semantics of pauvre (pitiful + N/N + not rich) clearly separates the uses. To sum up, the problem of adjective alternation does not depend on general principles, it is tightly linked to the item and to the NP within which it appears. Our study shows that the constraints previously proposed play a role in the placement of adjectives, but on a more specific level than the broad NP. In other words, locutors speak according to more specific patterns present in their linguistic knowledge.
In this paper we present an experimental toolbox for automatic tree-to-tree alignment based on local classification and alignment inference. The aligner implements a recurrent architecture for structural prediction using history features and a sequential classification procedure. The discriminative base classifier uses a log-linear model which enables simple integration of various features extracted from the data. The Lingua-Align toolbox provides a flexible framework for feature extraction including contextual properties and implements several alignment inference procedures. Various settings and constraints can be controlled via a simple frontend or called from external scripts. Lingua-Align supports different treebank formats and includes additional tools for conversion and evaluation. In our experiments we can show that our tree aligner produces results with high quality and outperforms unsupervised techniques proposed otherwise. It also integrates well with another existing tool for manual tree alignment which makes it possible to quickly integrate additional training material and to run semi-automatic alignment strategies. 1.
Treebank is a text corpus with syntactic annotation. It records the syntactic tree, i.e. the syntactic parse, of every sentence in running texts. Since 1990s, automatic parsing of natural languages has again become the focus of the international community of computational linguistics, and one of the crucial reasons is the realization of the Penn Treebank (PTB). The performances of statistical parsers, which are based on automatically induced Probabilistic Context-Free Grammar (PCFG), outperform significantly those rule-and unification-based parsers. A Treebank of any language in the world, represented with either phrase structures or dependency structures, takes sentence as its basic description unit. Dependency Grammar is a lexicalized grammar; it denies the notion of phrase structures and describes only the various word-word relations in a sentence, in which the head-word is the dominant of a given relation, and the other word of the word-pair at stake is the dependent of the head. Dependency Grammar creates a transparent interface between the dependency syntax and semantics of a language. This paper highly estimates the life force of the sentence-based syntax and the head-driven sentence analyzing method advocated by Jinxi Li, because they have not only dominated grammar teaching in middle schools more than half century before and after the foundation of the People’s Republic of China, but also guides the treebanking practice today.
The aim of this article is to analyze the formation of Old English adverbs (A-Y) as retrieved from the lexical database of Old English Nerthus within the theoretical framework of the Layered Structure of the Word. Firstly, a critical review of the literature on Old English adverb formation is offered in order to emphasize the necessity of an exhaustive and theoretically up-to-date study that distinguishes clearly synchronic from diachronic aspects on the one hand, and inflectional from derivational aspects on the other. Secondly, an exhaustive analysis of the derivation of adverbs in Old English by means of different word-formation processes (zero derivation, conversion, affixation and compounding) is given. In the theoretical part, the conclusion reached is that conversion requires a Complex Word structure and that a distinction has to be drawn between syntactic exocentricity and morphological exocentricity.
This article discusses the treatment of collocations in the context of along-term project on the development of multilingual NLP tools. Besides“classical” two-word collocations, we will focus on the case of complexcollocations (3 words or more) for which a recursive design is presented in theform of collocation of collocations. Although comparatively less numerous thantwo-word collocations, the complex collocations pose important challenges forNLP. The article discusses how these collocations are retrieved from corpora,inserted and stored in a lexical database, how the parser uses such knowledgeand what are the advantages offered by a recursive approach to complexcollocations.
uni-tuebingen.de This paper describes a CoNLL-style chunk representation for the Tübingen Treebank of Written German, which assumes a flat chunk structure so that each word belongs to at most one chunk. For German, such a chunk definition causes problems in cases of complex prenominal modification. We introduce a flat annotation that can handle these structures via a stranded noun chunk. 1
The article analyzes 97 elementary schoolbooks in Buenos Aires to determine which social representations about linguistic norm underlie in these school materials. The paper reviews -especially in the defi nitions of categories, and exercises and activities- the concepts of linguistic variety, standard language and español neutro. Based on these variables, this article sees the possible repercussions in social representations that students and teachers can develop from point of view of the publishing companies.
This paper describes the transfer component of a syntax-based Example-based Machine Translation system. The source sentence parse tree is matched in a bottom-up fashion with the source language side of a parallel example treebank, which results in a target forest which is sent to the target language generation component. The results on a 500 sentences test set are compared with a top-down approach to transfer of the same system, with the bottom-up approach yielding much better results. 1
We describe a process for converting the Penn Arabic Treebank into the CCG formalism. Previous efforts have yielded CCGbanks in English, German, and Turkish, thus opening these languages to the sophisticated computational tools developed for CCG and enabling further cross-linguistic development. Conversion from a context free grammar treebank to a CCGbank is a four stage process: head finding, argument classification, binarization, and category conversion. In the process of implementing a basic CCGbank conversion algorithm, we reveal properties of Arabic grammar that interfere with conversion, such as subject topicalization, genitive constructions, relative clauses, and optional pronominal subjects. All of these problematic phenomena can be resolved in a variety of ways- we discuss advantages and disadvantages of each in their respective sections. We detail these and describe our categorial analysis of each of these Arabic grammatical phenomena in depth, as well as technical details on their integration into the conversion algorithm. 1.
For centuries, scholars have explored the deep links among human languages. In this paper, we present a class of probabilistic models that use these links as a form of naturally occurring supervision. These models allow us to substantially improve performance for core text processing tasks, such as morphological segmentation, part-of-speech tagging, and syntactic parsing. Besides these traditional NLP tasks, we also present a multilingual model for the computational decipherment of lost languages. 1. Overview Electronic text is currently being produced at a vast and unprecedented scale across the languages of the world. Natural Language Processing (NLP) holds out the promise of automatically analyzing this growing body of text. However, over the last several decades, NLP research efforts have focused on the English language, often neglecting the thousands of other languages of the world (Bender, 2009). Most of these languages are currently beyond the reach of NLP technology due to several factors. One of these is simply the lack of the kinds of hand-annotated linguistic resources that have helped propel the performance of English language systems. For complex tasks of linguistic analysis, hand-annotated corpora can be prohibitively time-consuming and expensive to produce. For example, the most widely used annotated corpus in the English language, the Penn Treebank (Marcus et al., 1994), took years for a team of professional linguists to produce. It is unrealistic to expect such resources to ever exist for the majority of the world’s languages.
Two of the main corpora available for training discourse relation classifiers are the RST Discourse Treebank (RST-DT) and the Penn Discourse Treebank (PDTB), which are both based on the Wall Street Journal corpus. Most recent work using discourse relation classifiers have employed fully-supervised methods on these corpora. However, certain discourse relations have little labeled data, causing low classification performance for their associated classes. In this paper, we attempt to tackle this problem by employing a semi-supervised method for discourse relation classification. The proposed method is based on the analysis of feature cooccurrences in unlabeled data. This information is then used as a basis to extend the feature vectors during training. The proposed method is evaluated on both RST-DT and PDTB, where it significantly outperformed baseline classifiers. We believe that the proposed method is a first step towards improving classification performance, particularly for discourse relations lacking annotated data.
In this methodological investigation, we examined the influence of cultural background on viewers' interpretations of visual stimuli and verbs elicited by these materials. French and Mandarin native speakers' interpretations of seventeen short movies, produced by French speakers, depicting various state-changing actions were collected by a 25-item cultural protocol. A slight difference in the familiarity rating of movies is found between French and Mandarin participants. We also found that Mandarin speakers used more general verbs when describing actions depicted by movies with low familiarity rating and children used more conventional forms with movies of higher familiarity. Hierarchical cluster analyses were conducted in selecting movies that were matched in action-interpretations by both language groups.
Abstract Experiences with racism and age negatively affect how Afro-Brazilians in Salvador and São Paulo rate democracy. Older cohorts are more likely to rate democracy high compared to younger cohorts who rate it as low. Respondents in Salvador tend to rate democracy lower than respondents in São Paulo. Moreover, interviews reveal that as citizens believe they are not accorded full rights, they do not agree that Brazil's political system is fully democratic. Studies examining democracy in Brazil and racial politics throughout the diaspora would benefit from examining racialized experiences of citizens, rather than simply including the demographic variable of race. It is these experiences that affect rating of democracy rather than ascribed notions of race.
This study explored the conceptual framework of dieticians' intentions to recommend functional food and the mediating role of consumption frequency. A web-based survey was designed using a self-administered questionnaire. A sample of Korean dieticians (N=233) responded to the questionnaire that included response efficacy, risk perception, consumption frequency, and recommendation intention for functional foods. A structural equation model was constructed to analyze the data. We found that response efficacy was positively related to frequency of consumption of functional foods and to recommendation intention. Consumption frequency also positively influenced recommendation intention. Risk perception had no direct influence on recommendation intention; however, the relationship was mediated completely by consumption frequency. Dieticians' consumption frequency and response efficacy were the crucial factors in recommending functional foods. Dieticians may perceive risks arising from the use of functional foods in general, but the perceived risks do not affect ratings describing dieticians' intentions to recommend them. The results also indicated that when dieticians more frequently consume functional foods, the expression of an intention to recommend functional foods may be controlled by the salience of past behaviors rather than by attitudes.
In this paper, we argue for and demonstrate the use of Prolog as a tool to query annotated corpora. We present a case study based on the German TüBa-D/Z Treebank to show that flexible and efficient corpus querying can be started with a minimal amount of effort. We end this paper with a brief discussion of performance, that suggests that the approach is both fast enough and scalable. 1
Objective: To investigate whether interviewer personality, sex or being of the same sex as the interviewee, and training account for variance between interviewers’ ratings in a medical student selection interview. Design, setting and participants: In 2006 and 2007, data were collected from cohorts of each year's interviewers (by survey) and interviewees (by interview) participating in a multiple mini-interview (MMI) process to select students for an undergraduate medical degree in Australia. MMI scores were analysed and, to account for the nested nature of the data, multilevel modelling was used. Main outcome measures: Interviewer ratings; variance in interviewee scores. Results: In 2006, 153 interviewers (94% response rate) and 268 interviewees (78%) participated in the study. In 2007, 139 interviewers (86%) and 238 interviewees (74%) participated. Interviewers with high levels of agreeableness gave higher interview ratings (correlation coefficient [r] = 0.26 in 2006; r = 0.24 in 2007) and, in 2007, those with high levels of neuroticism gave lower ratings (r = − 0.25). In 2006 but not 2007, female interviewers gave higher overall ratings to male and female interviewees (t = 2.99, P = 0.003 in 2006; t = 2.16, P = 0.03 in 2007) but interviewer and interviewee being of the same sex did not affect ratings in either year. The amount of variance in interviewee scores attributable to differences between interviewers ranged from 3.1% to 24.8%, with the mean variance reducing after skills-based training (20.2% to 7.0%; t = 4.42, P = 0.004). Conclusion: This study indicates that rating leniency is associated with personality and sex of interviewers, but the effect is small. Random allocation of interviewers, similar proportions of male and female interviewers across applicant interview groups, use of the MMI format, and skills-based interviewer training are all likely to reduce the effect of variance between interviewers.
Nivre’s method was improved by enhancing deterministic dependency parsing through application of a tree-based model. The model considers all words necessary for selection of parsing actions by including words in the form of trees. It chooses the most probable head candidate from among the trees and uses this candidate to select a parsing action. In an evaluation experiment using the Penn Treebank (WSJ section), the proposed model achieved higher accuracy than did previous deterministic models. Although the proposed model’s worst-case time complexity is O(n 2), the experimental results demonstrated an average parsing time not much slower than O(n). 1
This dissertation considers the language socialization of law students. One message that the law students encounter is that legal Swedish is an entirely new language. The main aim is to investigate what linguistic norms are conveyed to the students through the teachers’ comments on the students’ texts and through various forms of writing instructions. The material consists of student texts with teacher comments and documentation on various phases of instruction with a focus on writing. Teacher comments on texts written during the first year of the law programme are analyzed and categorized. The analysis stems from two models. The first model is based on different text levels, like formal conventions of writing, sentence construction, text structure, word choice and style, and content. The second model distinguishes different linguistic norms based on three layers: The first layer consists of written language norms in general language practice, the second of academic language norms and the third of norms that are specific to the use of legal language. The results show that word choice and style is the most common category for the teachers’ comments in the first term of the law programme and content is the most common in the second term (with word choice and style the second most common). Formal conventions of writing, sentence structure and different types of grammatical constructions are some of the things the teachers criticize. Surprisingly few of the teachers’ comments concern more overarching aspects such as text structure or the aim and genre of the text. Comments are made on local features in the text, but rarely on more global features. The teaching practice that the writing of law students belongs to entails, among other things, that the students’ texts are assessed anonymously for the sake of fairness. This means that there is not much opportunity for a student to discuss the text with the teacher who commented on and assessed it. The construction of the teachers’ text comments is particularly important when dialogue between student and teacher on the text draft and final version is not an integral part of instruction. The teachers’ written comments are usually brief and do not allow much space for a consideration of linguistic norms and text patterns, which reduces the opportunities for the teachers and the law programme to contribute to a deeper linguistic awareness in the law students.
In this paper, we propose two different language modeling approaches, namely skip trigram and across sentence boundary, to capture the long range dependencies. The skip trigram model is able to cover more predecessor words of the present word compared to the normal trigram while the same memory space is required. The across sentence boundary model uses the word distribution of the previous sentences to calculate the unigram probability which is applied as the emission probability in the word and the class model frameworks. Our experiments on the Penn Treebank [1] show that each of our proposed models and also their combination significantly outperform the baseline for both the word and the class models and their linear interpolation. The linear interpolation of the word and the class models with the proposed skip trigram and across sentence boundary models achieves 118.4 perplexity while the best state-of-the-art language model has a perplexity of 137.2 on the same dataset. 1.
OBJECTIVE: To explore the possible differences in subjective analysis of the emotional stimuli from the International Affective Picture System between elderly and young samples. METHOD: 187 elderly subjects ranked the International Affective Picture System images according to the directions from the Manual of Affective Ratings. Their scores were compared to those obtained from International Affective Picture System studies with young people. RESULT: There is an age-related difference in arousal and valence in the International Affective Picture System rating. The correlation between affective valence and arousal is strong, and negative for the elderly. The expected versus the observed frequency of International Affective Picture System images between elderly and young samples show a statistical difference. CONCLUSION: This study shows an inter-age statistical dichotomy in how elderly and young people subjectively evaluate International Affective Picture System images.
There often exist multiple corpora for the same natural language processing (NLP) tasks. However, such corpora are generally used independently due to distinctions in annotation standards. For the purpose of full use of readily available human annotations, it is significant to simultaneously utilize multiple corpora of different annotation standards. In this paper, we focus on the challenge of constituent syntactic parsing with treebanks of different annotations and propose a collaborative decoding (or co-decoding) approach to improve parsing accuracy by leveraging bracket structure consensus between multiple parsing decoders trained on individual treebanks. Experimental results show the effectiveness of the proposed approach, which outperforms stateof-the-art baselines, especially on long sentences. 1
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 very essence of data. The use of linguistic summaries provides tools for the verbalisation of data analysis (mining) results which, in addition to the more commonly used visualisation, e.g. via a graphical user interface, can contribute to an increased human consistency and ease of use, notably for supporting decision makers via the data-driven decision support system paradigm. Two new relevant aspects of the analysis are also outlined which were first initiated by the authors. First, following Kacprzyk and Zadrożny, it is further considered how linguistic data summarisation is closely related to some types of solutions used in natural language generation (NLG). This can make it possible to use more and more effective and efficient tools and techniques developed in NLG. Second, similar remarks are given on relations to systemic functional linguistics. Moreover, following Kacprzyk and Zadrożny, comments are given on an extremely relevant aspect of scalability of linguistic summarisation of data, using a new concept of a conceptual scalability.
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), 103-113. \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.
Query expansion is a widely studied technique for improving information retrieval effectiveness. In this paper we proposed a new query expansion technique using the comprehensive thesaurus WordNet and its semantic relatedness measure modules. Word sense disambiguation are performed on original query sentence, yielding the concept of each term in the query. Based on those recovered concepts, expanded query terms are generated from WordNet lexical database. The proposed method has been evaluated in document retrieval on the Web using query sentence. Our extensive experimental results demonstrate a 7% precision improvement over retrieval methods not employing query expansion techniques.
This paper reports results on grammatical induction for French. We investigate how to best train a parser on the French Treebank We compare, for French, a supervised lexicalized parsing algorithm with a semi-supervised unlexicalized algorithm We report the best results known to us on French statistical parsing, that we obtained with the semi-supervised learning algorithm. The reported experiments can give insights for the task of grammatical learning for a morphologically-rich language, with a relatively limited amount of training data, annotated with a rather flat structure.
In this paper, we present our in-progress research tasks for building lexical database of the verb valences in the Arabic Quran using FrameNet frames. We study the verbs in their context in the Quran, and compare that with matching frames and frame evoking verbs in the English FrameNet. We analyze the gaps and make appropriate amendments to the FrameNet by adding new frame elements and relations.
Syntactic parsing is a central problem and a challenge in the field of natural language processing. It attracts many studies and consequently there exists the effective parsers for several popular languages such as English and Chinese. For Vietnamese parsing, there have been a few studies focusing on this problem, these studies lack of applying modern techniques, and no popular parser has been released. This paper presents the first study on developing a Vietnamese wide coverage parser based on lexicalized probabilistic context free grammar (LPCFG) and using a standard parsed corpus (similar to Penn Treebank). In this paper the Bikel's parser is modified to analyze Vietnamese. We also provide a comparison based on investigating different parsing models and different linguistic features. The best configuration achieves around 78% of F-score. © 2009 IEEE. Index Keywords: Central problems; Experimental studies; F-score; Linguistic features; NAtural language processing; Parsed corpora; Probabilistic context free grammars; Syntactic parsing; Treebanks; Computational linguistics; Context free grammars; Knowledge engineering; Natural language processing systems; Query languages; Standardization; Systems engineering; Formal languages
Alternating Structure Optimization (ASO) is a recently proposed linear Multitask Learning algorithm. Although its effective has been verified in both semi-supervised as well as supervised methods, yet they necessitate taking external resource as a prerequisite. Therefore, feasibility of employing ASO to further improve the performance merely rests on the labeled data on hand proves to be a task deserving close scrutiny. Catering to this challenging while untapped problem, this paper presents a novel application of ASO to the subtask of Shallow Semantic Parsing: Chunking. Our experiments on Chinese Treebank 5.0 present promising result in chunk analysis, and the error rate is reduced by 5.72%, proposing a profound way to further improve the performance.