Papers reviewed and determined not to be word norm studies. Use the flag icon to report errors or suggest re-inclusion.
18625 papers
We present several algorithms for assigning heads in phrase structure trees, based on different linguistic intuitions on the role of heads in natural language syntax. Starting point of our approach is the observation that a head-annotated treebank defines a unique lexicalized tree substitution grammar. This allows us to go back and forth between the two representations, and define objective functions for the unsupervised learning of head assignments in terms of features of the implicit lexicalized tree grammars. We evaluate algorithms based on the match with gold standard head-annotations, and the comparative parsing accuracy of the lexicalized grammars they give rise to. On the first task, we approach the accuracy of hand-designed heuristics for English and inter-annotation-standard agreement for German. On the second task, the implied lexicalized grammars score 4% points higher on parsing accuracy than lexicalized grammars derived by commonly used heuristics.
We present in this paper an approach to assessing student paraphrases in the intelligent tutoring system iSTART. The approach is based on measuring the semantic similarity between a student paraphrase and a reference text, called the textbase. The semantic similarity is estimated using knowledge-based word relatedness measures. The relatedness measures rely on knowledge encoded in Word-Net, a lexical database of English. We also experiment with weighting words based on their importance. The word importance information was derived from an analysis of word distributions in 2,225,726 documents from Wikipedia. Performance is reported for 12 different models which resulted from combining 3 different relatedness measures, 2 word sense disambiguation methods, and 2 word-weighting schemes. Furthermore, comparisons are made to other approaches such as Latent Semantic Analysis and the Entailer.
System for querying annotated treebanks in PML format. The querying uses it own query language with graphical representation. It has two different implementations (SQL and Perl) and several clients (TrEd, browser-based, command line interface).
This paper presents a set of experiments performed on parsing the Basque Dependency Treebank. We have applied feature propagation to dependency parsing, experimenting the propagation of several morphosyntactic feature values. In the experiments we have used the output of a parser to enrich the input of a second parser. Both parsers have been generated by Maltparser, a freely data-driven dependency parser generator. The transformations, combined with the pseudoprojective graph transformation, obtain a LAS of 77.12% improving the best reported results for Basque.
Statistical Machine Translation (SMT) systems rely heavily on the quality of the phrase pairs induced from large amounts of training data. Apart from the widely used method of heuristic learning of n-gram phrase translations from word alignments, there are numerous methods for extracting these phrase pairs. One such class of approaches uses translation information encoded in parallel treebanks to extract phrase pairs. Work to date has demonstrated the usefulness of translation models induced from both constituency structure trees and dependency structure trees. Both syntactic annotations rely on the existence of natural language parsers for both the source and target languages. We depart from the norm by directly obtaining dependency parses from constituency structures using head percolation tables. The paper investigates the use of aligned chunks induced from percolated dependencies in French–English SMT and contrasts it with the aforementioned extracted phrases. \nWe observe that adding phrase pairs from any other method improves translation performance over the baseline n-gram-based system, percolated dependencies are a good substitute for parsed dependencies, and that supplementing with our novel head percolation-induced chunks shows a general trend toward improving all system types across two data sets up to a 5.26% relative increase in BLEU.
A known problem of WordNet is that it is too ne-grained in its sense denitions. For instance, it does not distinguish between homographs and polysemes. This distinction is crucial in many natural language processing tasks. In this paper we propose to distinguish only between homographs withinWordNet data while merging all polysemous senses. The ultimate goal of this exercise is to compute a more coarsegrained version of linguistic database. In order to achieve this task we propose to merge all polysemous senses according to similarity scores computed by a hybrid algorithm. The key idea of the algorithm is to combine the similarity scores produced by diverse semantic similarity algorithms. We implemented the algorithm and evaluated it on the dataset extracted from the WordNet. The evaluation results are promising in comparison to the other state of the art approaches.
Bond ratings on state‐issued debt provide a signal to credit markets that help them charge an appropriate interest rate, based on the risk of payment default. Though actual default may occur only in extreme circumstances, observed differences in ratings and interest costs across states and time demonstrate that a sound economy, strong financials, and stable policies matter. When data on the factors that presumably affect ratings is public and easily accessible, making sense of differences of opinion between bond rating agencies is difficult. We suggest that such differences—observed as so‐called split bond ratings—are often ephemeral. Utilizing a simulation method to uncover the latent credit risk presented by each state, we show that split ratings on state bonds are often due to the fact that presumed category overlap between rating agencies is absent when evaluated on a common latent scale. Most observed state bond rating splits from 1997 through 2006 can be explained by this category mismatch. Our approach has broad implications for pricing state debt, as well as pricing rated debt in other capital market sectors.
The paper concentrates on obtaining hidden relationships among individual clauses of complex sentences from the Prague Dependency Treebank. The treebank contains only an information about mutual relationships among individual tokens (words, punctuation marks), not about more complex units (clauses). For the experiments with clauses and their parts (segments) it was therefore necessary to develop an automatic method transforming the original annotation into a scheme describing the syntactic relationships between clauses. The task was complicated by a certain degree of inconsistency in original annotation with regard to clauses and their structure. The paper describes the algorithm of deriving clause-related information from the existing annotation and its evaluation.
Full-Parsing systems able to analyze sentences robustly and completely at an appropriate accuracy can be useful in many computer applications like information retrieval and machine translation systems. Increasing the domain of locality by using tree-adjoining-grammars (TAG) caused some researchers to use it as a modeling formalism in their language application. But parsing with a rich grammar like TAG faces two main obstacles: low parsing speed and a lot of ambiguous syntactical parses. In order to decrease the parse time and these ambiguities, we use an idea of combining statistical chunker based on TAG formalism, with a heuristically rule-based search method to achieve the full parses. The partial parses induced from statistical chunker are basically resulted from a system named supertagger, and are followed by two different phases: error detection and error correction, which in each phase, different completion heuristics apply on the partial parses. The experiments on Penn Treebank show that by using a trained probability model considerable improvement in full-parsing rate is achieved.
Purpose The purpose of this paper is to investigate the characteristics of social network comments to give a broad overview to serve as a baseline for future research. Design/methodology/approach English comments from a representative sample of public MySpace profiles were examined with a collection of exploratory analyses, using automatic data processing, quantitative techniques and content analyses. Findings Comments were normally for general friendship maintenance and were typically short, with 95 per cent having 57 or fewer words. They contained a combination of standard spelling, apparently accidental mistakes, slang, sentence fragments, “typographic slang” and interjections. Several new creative spelling variants derived from previous forms of computer‐mediated communication have become extremely common, including u, ur,:), haha and lol. The vast majority of comments (97 per cent) contained at least one non‐standard language feature, suggesting that members almost universally recognise the informal nature of this kind of messaging. Research limitations/implications The investigation only covered MySpace and only analysed English comments. Practical implications MySpace comments should not be written in, or judged by, standard linguistic norms and may cause special problems for information retrieval. Originality/value This is the first large‐scale study of language in social network comments.
In this paper, we present a technique for visual analysis of documents based on the semantic representation of text in the form of a directed graph, referred to as semantic graph. This approach can aid data mining tasks, such as exploratory data analysis, data description and summarization. In order to derive the semantic graph, we take advantage of natural language processing, and carry out a series of operations comprising a pipeline, as follows. Firstly, named entities are identified and co-reference resolution is performed; moreover, pronominal anaphors are resolved for a subset of pronouns. Secondly, subject -- predicate -- object triplets are automatically extracted from the Penn Treebank parse tree obtained for each sentence in the document. The triplets are further enhanced by linking them to their corresponding co-referenced named entity, as well as attaching the associated WordNet synset, where available. Thus we obtain a semantic directed graph composed of connected triplets. The document's semantic graph is a starting point for automatically generating the document summary. The model for summary generation is obtained by machine learning, where the features are extracted from the semantic graph structure and content. The summary also has an associated semantic representation. The size of the semantic graph, as well as the summary length can be manually adjusted for an enhanced visual analysis. We also show how to employ the proposed technique for the Visual Analytics challenge.
Although body image is often conceptualized as a trait, there is growing recognition of situational influences. It was hypothesized that college women whose body image was more reactive to situational influences would be more likely to experience adverse weight-related outcomes, including obesity and binge eating behavior. In a pilot study, vignettes were rated to yield a range of situational body image ratings, and the three vignettes that captured the full range (lowest, median, and highest ratings) were used in the main study. In the main study, 138 female undergraduates (M = 19.5 yr., SD = 1.9; 74.6% "White or Caucasian") completed a survey that included the Questionnaire on Eating and Weight Patterns-Revised. They also provided Body Image State Scale ratings in response to three vignettes drawn from the pilot study. Body image "reactivity" was defined as the range over which body image scores varied for a given participant (standard deviation of the means). Contrary to predictions, overall body dissatisfaction across situations--but not situational reactivity--was related to higher BMI, greater likelihood of binge eating, and higher depression. Results suggest when body image dissatisfaction is relatively impervious to situational influences, it may be a risk factor for disordered eating behavior, whereas body image dissatisfaction that is responsive to situational influences appears to be less pathognomonic.
The Semantic Web was designed to unambiguously define and use ontologies to encode data and knowledge on the Web. Many people find it difficult, however, to write complex RDF statements and queries because doing so requires familiarity with the appropriate ontologies and the terms they define. We describe a system that automatically maps a set of ordinary English words to a set of appropriate ontology terms on the Semantic Web. We use the Swoogle Semantic Web search engine to provide ontology terms and ontology correlation statistics, the WordNet lexical database to resolve synonyms, and a practical three step approach to find the most suitable ontology context as well as appropriate ontology terms.
Chinese maximal-length phrases(maximal-length noun phrases and prepositional phrases) possess remarkable linguistic properties.Bidirectional labeling results of Chinese maximal-length phrases obtained using sequential classifiers reveal complementary properties in both directions.In this paper,both left-right and right-left sequential labeling were employed to identify the Chinese maximal-length noun phrases and prepositional phrases.Then a novel fork position based probabilistic algorithm was developed to fuse the bidirectional results.Experiments were carried out on the Penn Chinese Treebank,a segmented,part-of-speech tagged,and fully bracketed corpus.The results confirmed that the proposed algorithm is able to effectively exploit the complementary strengths of the two directions.
The Arabic language has a very rich/complex morphology. Each Arabic word is composed of zero or more prefixes, one stem and zero or more suffixes. Consequently, the Arabic data is sparse compared to other languages such as English, and it is necessary to conduct word segmentation before any natural language processing task. Therefore, the word-segmentation step is worth a deeper study since it is a preprocessing step which shall have a significant impact on all the steps coming afterward. In this article, we present an Arabic mention detection system that has very competitive results in the recent Automatic Content Extraction (ACE) evaluation campaign. We investigate the impact of different segmentation schemes on Arabic mention detection systems and we show how these systems may benefit from more than one segmentation scheme. We report the performance of several mention detection models using different kinds of possible and known segmentation schemes for Arabic text: punctuation separation, Arabic Treebank, and morphological and character-level segmentations. We show that the combination of competitive segmentation styles leads to a better performance. Results indicate a statistically significant improvement when Arabic Treebank and morphological segmentations are combined.
Dependency parsing has become increasingly popular for a surge of interest lately for applications such as machine translation and question answering. Currently, several supervised learning methods can be used for training high-performance dependency parsers if sufficient labeled data are available. However, currently used statistical dependency parsers provide poor results for words separated by long distances. In order to solve this problem, this article presents an effective dependency parsing approach of incorporating short dependency information from unlabeled data. The unlabeled data is automatically parsed by using a deterministic dependency parser, which exhibits a relatively high performance for short dependencies between words. We then train another parser that uses the information on short dependency relations extracted from the output of the first parser. The proposed approach achieves an unlabeled attachment score of 86.52%, an absolute 1.24% improvement over the baseline system on the Chinese Treebank data set. The results indicate that the proposed approach improves the parsing performance for longer distance words.
Despite the numerous bad press received by machine translation systems, they are invaluable in aiding users to quickly gather the essence of texts written in an unfamiliar language, particularly on the Web. Sufficiently accurate translation of lexical items is essential to satisfy such needs. Lexicon resources, especially a well-designed multilingual lexicon is needed to facilitate effective translation selection of lexical items, consisting of either single or multiple words. We describe the issues facing multilingual lexicography, and review how they are handled in selected multilingual lexical database projects. This preliminary surveys aims at identifying multilingual issues and concerns in translating lexical items from a machine translation perspective, and will guide the design of a multilingual lexicon intended for use in a machine translation system.
In this paper we extend a shallow parser [6] with prepositional phrase attachment. Although the PP attachment task is a well-studied task in a discriminative learning context, it is mostly addressed in the context of artificial situations like the quadruple classification task [18] in which only two possible attachment sites, each time a noun or a verb, are possible. In this paper we provide a method to evaluate the task in a more natural situation, making it possible to compare the approach to full statistical parsing approaches. First, we show how to extract anchor-pp pairs from parse trees in the GENIA and WSJ treebanks. Next, we discuss the extension of the shallow parser with a PP-attacher. We compare the PP attachment module with a statistical full parsing approach [4] and analyze the results. More specifically, we investigate the domain adaptation properties of both approaches (in this case domain shifts between journalistic and medical language). Keywords prepositional phrase attachment, shallow parsing, machine learning of language 1
In this paper semantic classes of Czech verbs are presented as they are obtained from the lexical database VerbaLex that has recently been built at the NLP Centre FI MU. At the moment we have in VerbaLex 82 semantic classes covering 10,482 Czech verb lemmata and 19,556 verb valency frames. We discuss the criteria for establishing semantic classes: the most important one is grouping verbs according to their senses. The second one exploits relations between semantic classes of Czech verbs and semantic roles and subcategorization features as they are used in VerbaLex valency frames. We also touch on the issue of the ontology that could be used to describe the meanings of the verbs in the semantic classes. The semantic classification of Czech verbs can be extended for other languages via Interlingual Index (ILI) existing in WordNets and it can be used in the various applications in the NLP area (machine translation, syntactic analysis, semantic search, information extraction and others).
WordNet, a lexical database for English, is organized around semantic and lexical relationships between synsets, concepts represented by sets of synonymous word senses. Offering reasonably comprehensive coverage of the nouns, verbs, adjectives, and adverbs of general English, WordNet is a widely used resource for dealing with the ambiguity that arises from homonymy, polysemy, and synonymy. WordNet is used in many information-related tasks and applications (e.g., word sense disambiguation, semantic similarity, lexical chaining, alignment of parallel corpora, text segmentation, sentiment and subjectivity analysis, text classification, information retrieval, text summarization, question answering, information extraction, and machine translation).
We present a semi-supervised method to improve statistical parsing performance. We focus on the well-known problem of lexical data sparseness and present experiments of word clustering prior to parsing. We use a combination of lexicon-aided morphological clustering that preserves tagging ambiguity, and unsupervised word clustering, trained on a large unannotated corpus. We apply these clusterings to the French Treebank, and we train a parser with the PCFG-LA unlexicalized algorithm of (Petrov et al., 2006). We find a gain in French parsing performance: from a baseline of F1=86.76% to F1=87.37% using morphological clustering, and up to F1=88.29% using further unsupervised clustering. This is the best known score for French probabilistic parsing. These preliminary results are encouraging for statistically parsing morphologically rich languages, and languages with small amount of annotated data.
The process of marking up the syntactic structure of the sentences in a corpus is facilitated by having a graphical tree editor. A type specification constraints the nature of acceptable trees, and visual indication of non-conformities makes corrections easy. Projectivity checking can also be a useful way to notice errors, depending on the linguistic design of the tree type specification. A small test corpus has been marked up using this method. Only the English language has been investigated so far, and the tool takes no other input than the type specification and the corpus. Adding a lexical database is an obvious way to make it more generally useful.
Collocations constitute a subclass of multi-word expressions that are particularly problematic for machine translation, due 1) to their omnipresence in texts, and 2) to their morpho-syntactic properties, allowing virtually unlimited variation and leading to long-distance dependencies.Since existing MT systems incorporate mostly local information, these are arguably ill-suited for handling those collocations whose items are not found in close proximity.In this article, we describe an integrated environment in which collocations (and possibly their translation equivalents) are first identified from text corpora and stored in the lexical database of a translation system, then they are employed by this system, which is capable of dealing with syntactic transformations as it is based on a deep linguistic approach.We compare the performance of our system (in terms of collocation translation adequacy) with that of two major MT systems, one statistical, and the other rule-based.Our results confirm that syntactic variation affects translation quality and show that a deep syntactic approach is more robust in this sense, especially for languages with freer word order (e.g., German) and richer morphology (e.g., Italian) than English.
For Arabic, diacritizing written text is important for many NLP tasks. In the work presented here, we investigate the quality of a diacritization approach, with a high success rate for treebank data but with a more limited success on realworld data. One of the problems we encountered is the non-standard use of the hamza diacritic, which leads to a decrease in diacritization accuracy. If an automatic hamza restoration module precedes diacritization, the results improve from a word error rate of 9.20% to 7.38% in treebank data, and from 7.96% to 5.93% on selected real-world texts. This shows clearly that hamza restoration is a necessary step for improving diacritization quality for Arabic real-world texts.
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
The paper reports on the main findings of the LANCHART language attitudes studies. These studies were designed to falsify (or modify) the picture of adolescent language ideology – and its role in language change – that had emerged from previous sociolinguistic studies in Denmark. This picture is formulated as three hypotheses: (1) There are two value systems at two levels of consciousness, (2) Language change is governed by subconscious values, (3) Copenhagen is Denmark’s only linguistic norm centre. Following strict guidelines for data collection among 9th graders (aged 15–16) in Copenhagen, Næstved, Vissenbjerg, Odder, and Vinderup we obtained subconsciously offered attitudes that could be compared with consciously offered attitudes. The results neither falsify nor modify the established picture but strongly confirm it.
ABSTRACT: This paper argues that the ‘world Englishes paradigm’ and English as a lingua franca (ELF) research, despite important differences, have much in common. Both share the pluricentric assumption that ‘English’ belongs to all those who use it, and both are concerned with the sociolinguistic, socio‐psychological, and applied linguistic implications of this assumption. For example, issues of language contact, variation and change, linguistic norms and their acceptance, ownership of the language, and expression of social identities are central to both WE and ELF research. The growing body of descriptive ELF research that is now becoming available can thus add substance to work in the field as a whole. It can also offer fresh perspectives on several theoretical constructs central to WE, such as ‘community’, ‘variety’, ‘lingua franca’, even ‘language’.
In this paper we describe an approach to target language modeling which is based on a large treebank. We assume a bag of bags as input for the target language gener-ation component, leaving it up to this com-ponent to decide upon word and phrase or-der. An experiment with Dutch as target language shows that this approach to can-didate translation reranking outperforms standard n-gram modeling, when measur-ing output quality with BLEU, NIST, and TER metrics. 1
Manually annotated corpora are valuable but scarce resources, yet for many annotation tasks such as treebanking and sequence labeling there exist multiple corpora with different and incompatible annotation guidelines or standards. This seems to be a great waste of human efforts, and it would be nice to automatically adapt one annotation standard to another. We present a simple yet effective strategy that transfers knowledge from a differently annotated corpus to the corpus with desired annotation. We test the efficacy of this method in the context of Chinese word segmentation and part-of-speech tagging, where no segmentation and POS tagging standards are widely accepted due to the lack of morphology in Chinese. Experiments show that adaptation from the much larger People's Daily corpus to the smaller but more popular Penn Chinese Treebank results in significant improvements in both segmentation and tagging accuracies (with error reductions of 30.2% and 14%, respectively), which in turn helps improve Chinese parsing accuracy.
In this paper, I consider whether tenses, temporal indexicals, and other indexicals are contextually dependent on the context of assessment (or a-contextual), rather than, as is usually thought, contextually dependent on the context of utterance (u-contextual). I begin by contrasting two possible linguistic norms, governing our use of context sensitive expressions, especially tenses and temporal indexicals (§§2 and 3), and argue that one of these norms would make those expressions u-contextual, while the other would make them a-contextual (§4). I then ask which of these two norms are followed by English speakers (§5). Finally, I argue that the existence of a-contextuality does not in any sense entail ‘relativism’ about truth (§6).
This paper presents a simple and effective approach to improve dependency parsing by using subtrees from auto-parsed data. First, we use a baseline parser to parse large-scale unannotated data. Then we extract subtrees from dependency parse trees in the auto-parsed data. Finally, we construct new subtree-based features for parsing algorithms. To demonstrate the effectiveness of our proposed approach, we present the experimental results on the English Penn Treebank and the Chinese Penn Treebank. These results show that our approach significantly outperforms baseline systems. And, it achieves the best accuracy for the Chinese data and an accuracy which is competitive with the best known systems for the English data.
In the Computational Linguistics community, much work is put into the creation of large, high-quality linguistic resources, often with complex annotation. In order to make these resources accessible to non-technical audiences, formalisms for searching and filtering are needed, like the TIGER corpus query language. Recently, augmented treebanks have been published, including the SALSA corpus which features frame semantic annotation on top of syntactic structure. We design an extension for the TIGER language which allows searching for frame structures along with syntactic annotation. To achieve this, the TIGER ob ject model is expanded to include frame semantics, while remaining fully backwards-compatible and add these extensions to our own implementation of TIGER.
Impersonating Identity in Spaces of Difference is an ALTERnative discourse of dis/ruption, decolonization, deconstruction....Writing on the b/orders of theories, disciplines, genres, cultures....this re/search weaves together personal, familial, and societal stories of silence and silencings. By subverting conventional academic texts and hegemonic frameworks of Canadian "MultiCULTural institutions and Canadian "MultiCULTural society," this subaltern re/search claims a space for marginalized voices. Conventional ways of knowing, being, becoming are generatively disRupted in order to create awareness of the continuing legacies of colonialism, modernity, patriarchy...and to highlight the urgent need to provide genuine spaces of "belonging" that inclusively honour and respect the gifts of all individuals. Entangled in the in/visible hierarchical realities of "Canadianness," this in/quiry articulates decolonizing resistance. The layers of the impure academic textual body perform the multiple fragmented intertextual layers of the improper Indocanadian/ Can-indian body creating em(bodi)ed re-imag(e)inings of epistemology and pedagogy. This tense, restless landscape of multi-languages, multi-genres, multimeanings, multi-truths is an inter-Ruption of predetermined b/orders and predetermined bodies of predetermined purity, in a world of ever-changing multiplicity. Drawing on "difference" in multiCULTuralism, language, voice, and identity, this performative work travels in and out of questions of absence, "hybridity," foreignness, loss, displacement, marginalization, patriarchy, colonialism, modernity....In this powerful and liberating form of non-traditional in/quiry, meaning making takes precedence over conventional stylistic or pre-established structures and acknowledges personal "ethnic" experience as a valuable form of reliable "academic" knowledge. This trans-disciplinary, transformative, transcultural...in/quiry disRupts traditional hegemonic narratives and challenges the conventional notions of re/search and writing through its form and content. Through the braided weaving of English, French and Punjabi, personal stories, familial narratives, prose, letters, e-mails, cross cultural conversations, visual imagery, historical documents, "subtexts," "surtexts," intertexts, collaborative texts...undermine and upset hegemonic linguistic norms. Fiction, fantasy, History, herstory, theirstories, memory...are juxtaposed through mixed non-linear genres and codes in protest of violent acts of com(form)ity, exclusion and censorship. Stories of India and Canada find themselves interwoven unexpectedly, betraying the lies and the truths of patriarchy, colonialism, modernity, multiCULTuralism, transCULTuralisms...disrupting clean, linear readings of writing and of research. This experiment with nonstructure and typography attempts to actively decolonize, deconstruct/re-construct imposed academic and social identities in a more meaningful way and provide readers with a sense of living in the "transculturality" of the "diasporic in-between." Only by provoking a critical, cross-cultural "INNERstanding" of History, POWer, systemic marginalization, colonialism and modernity can we begin to relationally take on the collective response-abilty for social justice in policy and practice ~ in the word and in the wor(l)d.
We present a cost effective strategy for the creation of a mid-size fine-grained dependency treebank of surface- and deep-syntactic structures as defined in the Meaning-Text Theory for Spanish. The strategy starts from a small seed dependency corpus, the AnCora corpus, whose annotation is considerably more coarse-grained than our target annotation. We show that this discrepancy can be bridged largely by automatic means, relying upon contextual information and leaving thus minimal work to the annotators. This allows us to develop the resources with limited human effort within a limited period of time. We also propose a preliminary evaluation of the actual amount of work that the annotation process requires. 1
Az eladasban a Szeged Treebank fuggsegi fa formatumra torten atalakitasanak folyamatat mutatjuk be. Az eredetileg frazisstrukturalt treebankbl automatikus konverzio eredmenyekeppen letrejott fuggsegi fakat kezi uton ellenriztuk es javitottuk, letrehozva ezzel az els magyar nyelv kezzel annotalt dependenciakorpuszt. Jelenleg az uzleti hireket, ujsaghireket es jogi szovegeket tartalmazo alkorpuszok annotacioja fejezdott be, de terveink kozott szerepel a teljes korpusz atalakitasa fuggsegi fa formatumra. Az elkeszult adatbazis hasznosithato tobbek kozott az informaciokinyeresben es a gepi forditasban is.
In this paper, we would like to introduce a new approach to recover Vietnamese text's accents. Given a Vietnamese text in which accents are lost, our goal is to seek for a recovered text that yields a best lexical probability. Using a dynamic programming approach, we first build a model of language for Vietnamese as a lexical database which gives lexical probabilities to Vietnamese sentences. Second, we construct a map of literal translations of Vietnamese words to restrict our searching space. Finally, we apply dynamic programming as a searching engine to seek out the most probable sentence. We also use the co-occurrence graph to increase the accuracy of selection, the experimental results show that the average accuracy of our approach is about 93%-94%.
GLARF relations are generated from treebank and parses for English, Chinese and Japanese. Our evaluation of system output for these input types requires consideration of multiple correct answers.