Papers reviewed and determined not to be word norm studies. Use the flag icon to report errors or suggest re-inclusion.
18625 papers
For scientists and researchers, it is very critical to ensure knowledge is accessible for re-use and development. Moreover, the way we store and manage scientific articles and their metadata in digital libraries determines the amount of relevant articles we can discover and access depending on what is actually meant in a search query. Yet, are we able to explore all semantically relevant scientific documents with the existing keyword-based search information retrieval systems? This is the primary question addressed in this thesis. Hence, the main purpose of our work is to broaden or expand the knowledge spectrum of researchers working in an interdisciplinary domain when they use the information retrieval systems of multidisciplinary digital libraries. However, the problem raises when such researchers use community-dependent search keywords while other scientific names given to relevant concepts are being used in a different research community.Towards proposing a solution to this semantic exploration task in multidisciplinary digital libraries, we applied several text mining approaches. First, we studied the semantic representation of words, sentences, paragraphs and documents for better semantic similarity estimation. In addition, we utilized the semantic information of words in lexical databases and knowledge graphs in order to enhance our semantic approach. Furthermore, the thesis presents a couple of use-case implementations of our proposed model
Neural networks have shown great potential in language modeling. Currently, the dominant approach to language modeling is based on recurrent neural networks (RNNs) and convolutional neural networks (CNNs). Nonetheless, it is not clear why RNNs and CNNs are suitable for the language modeling task since these neural models are lack of interpretability. The goal of this paper is to tailor an interpretable neural model as an alternative to RNNs and CNNs for the language modeling task. This paper proposes a unified framework for language modeling, which can partly interpret the rationales behind existing language models (LMs). Based on the proposed framework, an interpretable neural language model (INLM) is proposed, including a tailored architectural structure and a tailored learning method for the language modeling task. The proposed INLM can be approximated as a parameterized auto-regressive moving average model and provides interpretability in two aspects: component interpretability and prediction interpretability. Experiments demonstrate that the proposed INLM outperforms some typical neural LMs on several language modeling datasets and on the switchboard speech recognition task. Further experiments also show that the proposed INLM is competitive with the state-of-the-art long short-term memory LMs on the Penn Treebank and WikiText-2 datasets.
Most current language modeling techniques only exploit co-occurrence, semantic and syntactic information from the sequence of words. However, a range of information such as the state of the speaker and dynamics of the interaction might be useful. In this work we derive motivation from psycholinguistics and propose the addition of behavioral information into the context of language modeling. We propose the augmentation of language models with an additional module which analyzes the behavioral state of the current context. This behavioral information is used to gate the outputs of the language model before the final word prediction output. We show that the addition of behavioral context in language models achieves lower perplexities on behavior-rich datasets. We also confirm the validity of the proposed models on a variety of model architectures and improve on previous state-of-the-art models with generic domain Penn Treebank Corpus.
OBJECTIVE: Facial expressions communicate emotional states and regulate social bonds. An approach or avoidance-based valence might interact with direct or averted gaze to elicit different attentional allocation. These processes might be aberrant in major depression or first-episode psychosis and this requires empirical investigation. METHOD: This study examined higher order, controlled attentional processing of emotional facial expressions (happy, neutral, angry and fearful), with direct or averted gaze, using electroencephalogram (EEG) measures of the face-elicited Late Positive Potential (LPP), in young people diagnosed with major depression or first-episode psychosis, compared with a healthy control group. RESULTS: In the control group, there was no evidence of increased attentional allocation to emotional facial expressions, or to facial expressions with a matching emotional expression and gaze direction. There was no evidence, in the depression or first-episode psychosis groups, for a threat-based, attentional hypersensitivity to fearful or angry facial expressions, nor for this effect to be potentiated in response to angry direct or fearful averted gaze faces. However, the absence of such effects could not be concluded due to sample size and the absence of stimulus arousal and valence ratings. Importantly, there was significantly increased attentional allocation in the first-episode psychosis group to facial expressions regardless of emotional expression or gaze direction, compared to both the depression and control group. CONCLUSIONS: There might be an attentional hypersensitivity to facial expressions regardless of emotional expression or gaze direction in first-episode psychosis.
Dependency parser is one of the most important fundamental tools in the natural language processing, which extracts structure of sentences and determines the relations between words based on the dependency grammar. The dependency parser is proper for free order languages, such as Persian. In this paper, data-driven dependency parser has been developed with the help of phrase-structure parser for Persian. The defined feature space in each parser is one of the important factors in its success. Our goal is to generate and extract appropriate features to dependency parsing of Persian sentences. To achieve this goal, new semantic and syntactic features have been defined and added to the MSTParser by stacking method. Semantic features are obtained by using word clustering algorithms based on syntagmatic analysis and syntactic features are obtained by using the Persian phrase-structure parser and have been used as bit-string. Experiments have been done on the Persian Dependency Treebank (PerDT) and the Uppsala Persian Dependency Treebank (UPDT). The results indicate that the definition of new features improves the performance of the dependency parser for the Persian. The achieved unlabeled attachment score for PerDT and UPDT are 89.17% and 88.96% respectively.
BACKGROUND: Cutting is the most common method of non-suicidal self-injury (NSSI) to reduce inner tension in patients with Borderline Personality Disorder (BPD). Aim of this study was to compare pain perception induced by an incision and by application of a surrogate model for sharp mechanical pain (a non-invasive "blade") in BPD. METHODS: 22 female patients and 20 healthy controls (HC) received a small incision into the volar forearm, a 7s-blade application on the same side, and non-invasive phasic stimuli (pinprick, blade, laser, tactile). Pain intensity as well as affective versus sensory components were assessed. RESULTS: Incision was rated similarly by both groups (BPD: 28.6 ± 5.5 vs. HC: 33.9 ± 6.6; mean maximum pain ± SEM; p > 0.8), without significant difference for "7-s-blade" (BPD: 18.1 ± 3.8 vs. HC: 25.3 ± 3.6; mean maximum pain ± SEM; p > 0.17) or between "7-s-blade" and incision (BPD: p > 0.12; HC: p > 0.84). However, patients' intensity ratings returned significantly faster to baseline after incision (BPD: 38.9 ± 12.6 s vs. HC: 74.52 ± 11.5 s; p < 0.05), and patients evaluated "blade" and incision without any affective and with different sensory descriptors, indicating an altered evaluation of NSSI-like stimulation with qualitative in addition to quantitative differences-especially for the sharp pain component. CONCLUSIONS: The reduced perception of suprathreshold nociceptive stimuli is based on a missing affective component and specific loss of the perception of "sharpness" as part of the sensory component of pain. The results further demonstrate the usefulness of the "blade" for the perception of sharpness in patients. SIGNIFICANCE: Patients with Borderline Personality Disorder (BPD) who engage in non-suicidal self-injury (NSSI) report less pain in response to phasic nociceptive stimuli. In comparing an invasive pain stimulus to phasic nociceptive stimuli in BPD patients, the "blade" as non-invasive surrogate model for sharp mechanical pain in psychiatric patients is used. In contrast to healthy volunteers, BPD patients do not report significant affective ratings and specifically display a reduced sensory component for sharpness.
We study how to leverage off-the-shelf visual and linguistic data to cope with out-of-vocabulary answers in visual question answering task. Existing large-scale visual datasets with annotations such as image class labels, bounding boxes and region descriptions are good sources for learning rich and diverse visual concepts. However, it is not straightforward how the visual concepts can be captured and transferred to visual question answering models due to missing link between question dependent answering models and visual data without question. We tackle this problem in two steps: 1) learning a task conditional visual classifier, which is capable of solving diverse question-specific visual recognition tasks, based on unsupervised task discovery and 2) transferring the task conditional visual classifier to visual question answering models. Specifically, we employ linguistic knowledge sources such as structured lexical database (e.g. WordNet) and visual descriptions for unsupervised task discovery, and transfer a learned task conditional visual classifier as an answering unit in a visual question answering model. We empirically show that the proposed algorithm generalizes to out-of-vocabulary answers successfully using the knowledge transferred from the visual dataset.
This paper describes a method of inflecting and linearizing a lemmatized dependency tree by: (1) determining a regular expression and substitution to describe each productive wordform rule; (2) learning the dependency distance tolerance for each head-dependent pair, resulting in an edge-weighted directed acyclic graph (DAG); and (3) topologically sorting the DAG into a surface realization based on edge weight. The method's output for 11 languages across 18 treebanks is competitive with the other submissions to the Second Multilingual Surface Realization Shared Task (SR'19).
The parataxis relation as defined for Universal Dependencies 2.0 is general and, for this reason, sometimes hard to distinguish from competing analyses, such as coordination, conj, or apposition, appos. The specific subtypes that are listed for parataxis are also quite different in character. In this study we first show that the actual practice by UD-annotators is varied, using the parallel UD (PUD-) treebanks as data. We then review the current definitions and guidelines and suggest improvements.
Supertagging is a sequence prediction task where each word is assigned a piece of complex syntactic structure called a supertag. We provide a novel approach to multi-task learning for Tree Adjoining Grammar (TAG) supertagging by deconstructing these complex supertags in order to define a set of related but auxiliary sequence prediction tasks. Our multi-task prediction framework is trained over the exactly same training data used to train the original supertagger where each auxiliary task provides an alternative view on the original prediction task. Our experimental results show that our multi-task approach significantly improves TAG supertagging with a new state-of-the-art accuracy score of 91.39% on the Penn treebank supertagging dataset.
Transition-based parsing of natural language uses transition systems to build directed annotation graphs (digraphs) for sentences. In this paper, we define, for an arbitrary ordered digraph, a unique decomposition and a corresponding linear encoding that are associated bijectively with each other via a new transition system. These results give us an efficient and succinct representation for digraphs and sets of digraphs. Based on the system and our analysis of its syntactic properties, we give structural bounds under which the set of encoded digraphs is restricted and becomes a context-free or a regular string language. The context-free restriction is essentially a superset of the encodings used previously to characterize properties of noncrossing digraphs and to solve maximal subgraphs problems. The regular restriction with a tight bound is shown to capture the Universal Dependencies v2.4 treebanks in linguistics.
Abstract We study the distribution of the nominal and copular construction of predicate nominals in a subset of authors from the Ancient Greek Dependency Treebank ( AGDT ). We concentrate on the texts of the historians Herodotus, Thucydides (both 5th century BCE) and Polybius (2nd century BCE). The data comprise a sample of 440 sentences (Hdt = 175, Thuc = 91, Pol = 174). We analyze the impact of four features that have been discussed in the literature and can be observed in the annotation of AGDT: (1) order of constituents, (2) part of speech of the subjects, (3) type of clause and (4) length of the clause. Furthermore, we test how the predictive power of these factors varies in time from Herodotus and Thucydides to Polybius with the help of a logistic-regression model. The analysis shows that, contrary to a simplistic opinion, the nominal construction does not drop into irrelevance in Hellenistic Greek. Moreover, an analysis of the distributions in the authors highlights a remarkable continuity in the usage patterns. Further work is needed to improve the predictive power of our logistic-regression model and to integrate more data in view of a more comprehensive quantitative diachronic study.
Different from the current syntax parsing based on deep learning, we present a novel Chinese parsing method, which is based on Sliding Match of Semantic String (SMOSS). (1) Training stage: In a treebank, headwords of tree nodes are represented by semantic codes given in the Synonym Dictionary (Tongyici Cilin). N-gram semantic templates are extracted from every layer of a syntax tree by means of sliding window to establish one N-gram semantic template library. (2) Parsing stage: Words of a sentence, including headwords of chunks, are represented by the semantic codes from Tongyici Cilin. With the sliding window method, N-gram semantic code strings are extracted to match with the templates in the N-gram semantic template library; subsequently, the mapping information of the matched templates is employed to guide the chunking of semantic code strings. The Chinese syntax parsing is completed through continuous matching and chunking. On the same training scale, N-gram semantic template can create favorable conditions for flexible matching and improve the syntax parsing performance. With train and test sets from the Tsinghua Chinese Treebank (TCT), the results are F1-score 99.71% (closed test) and F1-score 70.43% (open test), respectively.
WordNet is a lexical database for languages, the difference between WordNet and dictionaries in general is that WordNet focuses on the synonyms. The main unit of WordNet is synonym set (synset), synset is a set of one or more words that have the same meaning and certainly can be replaced in certain contexts. Synset is a very important element in implementing WordNet. In this paper, an analysis of the synonym extraction process is carried out by using commutative approach, the data test obtained from the Oxford Paperback Thesaurus by taking 51 word entries. Commutative method has similar characters with synonym set, synonym set can replace each other in certain contexts. The data test extraction process is carried out until the performance measurement evaluation process using F1Score. The system generates synonym sets that matched with the manual extraction, the result of F1Score between the program and Princeton synonym sets are worth 10%.
Gender role theory argues that men are socialized to adhere to masculine norms that foment resistance to help-seeking in general and medical care in particular. We hypothesize an automatic link in memory between medical care and masculinity among men, developed through gender role socialization, such that by the time men have reached adulthood, medical-related stimuli inescapably and unintentionally activate masculinity-relevant concepts. Across two studies we employ a lexical decision task with medical and non-medical primes to demonstrate an automatic link between medicine and masculinity (in addition to other concepts) among male university students. Further, we illustrate how self-affirmation may temporarily untether the two concepts. A novel “masculine-affirmation” was also explored, which increased the medicine–masculinity link.
In recent years, neural networks have been widely used for language modeling in different tasks of natural language processing. Results show that long short-term memory (LSTM) neural networks are appropriate for language modeling due to their ability to process long sequences. Furthermore, many studies are shown that extra information improve language models (LMs) performance. In this research, we propose parallel structures for incorporating part-of-speech tags into language modeling task using both the unidirectional and bidirectional type of LSTMs. Words and part-of-speech tags are given to the network as parallel inputs. In this way, to concatenate these two paths, two different structures are proposed according to the type of network used in the parallel part. We analyze the efficiency on Penn Treebank (PTB) dataset using perplexity measure. These two proposed structures show improvements in comparison to the baseline models. Not only does the bidirectional LSTM method gain the lowest perplexity, but it also has the lowest training parameters among our proposed methods. The perplexity of proposed structures has reduced 1.5% and %13 for unidirectional and bidirectional LSTMs, respectively.
This two-phased, sequential mixed-methods study investigates how raters are influenced by different rating scales on a college-level English as a second language (ESL) writing placement test. In Phase I, nine certified raters rated 152 essays using a holistic, profile-based scale; in Phase II, they rated 200 essays using a binary, analytic scale developed based on the holistic scale and 100 essays using both rating scales. Ratings were examined both quantitatively through Rasch modeling and qualitatively via think-aloud protocols and semi-structured interviews. Findings from Phase I revealed that, despite satisfactory internal consistency, the raters demonstrated relatively low rater agreement and individual differences in their use of the holistic scale. Findings from Phase II showed that the binary, analytic scale led to much improvement in rater consensus and rater consistency. Another finding from Phase II suggests that the binary, analytic scale helped the raters deconstruct the holistic scale, reducing their cognitive burden. This study represents a creative use of a binary, analytic scale to guide raters through a holistic rating scale. Implications regarding how a rating scale affects rating behavior and performance are discussed.
We recast dependency parsing as a sequence labeling problem, exploring several encodings of dependency trees as labels. While dependency parsing by means of sequence labeling had been attempted in existing work, results suggested that the technique was impractical. We show instead that with a conventional BiLSTM-based model it is possible to obtain fast and accurate parsers. These parsers are conceptually simple, not needing traditional parsing algorithms or auxiliary structures. However, experiments on the PTB and a sample of UD treebanks show that they provide a good speed-accuracy tradeoff, with results competitive with more complex approaches.
In this paper, we extend recent approaches to Lexicalized Tree Adjoining Grammar (LTAG) parsing that combine supertagging with dependency parsing. In other words, we assign supertags (= unanchored elementary trees) to lexical items and we compute substitution/adjunction arcs between them. Kasai et al. (2017, 2018) jointly predict these structures with a neural graph-based parser. Predicting 1-best supertags and dependency arcs (as in Kasai et al. (2017, 2018)) however leads only to partial parsing due to incompatibilities between elementary trees and derivation trees. We therefore extend the approach described in Kasai et al. (2017, 2018) to n-best supertags and k-best dependency arcs and combine it with a subsequent A*-parsing step that extends the TAG parser from Waszczuk (2017). We show that this architecture allows for efficient full TAG parsing while being sufficiently accurate. We test our architecture on an LTAG extracted from the French Treebank (FTB).
Abstract This chapter analyzes the interface between prosodic features and situational variables in the Rhapsodie treebank. First, we provide general quantitative information. Second, we present a preliminary set of statistical analyses performed on six specific prosodic variables, that provides two types of information. While Principal Component Analysis is useful to identify major trends in the data, inferential analysis is the first step in modeling the effect of the different modalities of each situational variable on the prosodic patterns, before considering predictive modeling. Finally, by focusing on the IPA unit, we present a third statistical method based on the computation of specificity indexes.
This dataset is a PostgreSQL dump of a database generated using autocode_big from the data in the LexiRumah lexical database of eastern Indonesia and Timor-Leste. The dataset contains the forms from the parent dataset, together with automatically generated sound correspondence scorers, pairwise similarity scores, cognate classes, and alignments, all created using Lingpy's LexStat algorithm.
This paper is centered around two main contributions: the first one consists in introducing several procedures for generating random dependency trees with constraints; we later use these artificial trees to compare their properties with the properties of natural trees (i.e trees extracted from treebanks) and analyze the relationships between these properties in natural and artificial settings in order to find out which relationships are formally constrained and which are linguistically motivated. We take into consideration five metrics: tree length, height, maximum arity, mean dependency distance and mean flux weight, and also look into the distribution of local configurations of nodes. This analysis is based on UD treebanks (version 2.3, Nivre et al. 2018) for four languages: Chinese, English, French and Japanese.
Cross-lingual dependency parsing involves transferring syntactic knowledge from one language to another. It is a crucial component for inducing dependency parsers in low-resource scenarios where no training data for a language exists. Using Faroese as the target language, we compare two approaches using annotation projection: first, projecting from multiple monolingual source models; second, projecting from a single polyglot model which is trained on the combination of all source languages. Furthermore, we reproduce multisource projection Finally, we apply multi-treebank modelling to the projected treebanks, in addition to or alternatively to polyglot modelling on the source side. We find that polyglot training on the source languages produces an overall trend of better results on the target language but the single best result for the target language is obtained by projecting from monolingual source parsing models and then training multi-treebank POS tagging and parsing models on the target side.
This paper describes the application of delexicalized cross-lingual parsing on Occitan with a view to building the first dependency treebank of this language. Occitan is a Romance language spoken in the south of France and in parts of Italy and Spain. It is a relatively low-resourced language and does not have a syntactically annotated corpus as of yet. In order to facilitate the manual annotation process, we train parsing models on the existing Romance corpora from the Universal Dependencies project and apply them to Occitan. Special attention is given to the effect of this cross-lingual annotation on the work of human annotators in terms of annotation speed and ease.
Fuzzy Semantic Similarity Measures are algorithms that are able to compare two or more short texts that contain human perception based words and return a numeric measure of similarity of meaning between them. Such similarity is computed using a weighting, comprised of the semantic and the syntactic composition of the short text. Similarities of individual words are computed through the use of a corpus, and ontological structures based on both WordNet - a well-known lexical database of English, and on category specific fuzzy ontologies created from the derivation of Type-I or Type-II interval fuzzy sets from human perceptions of fuzzy words. Currently, linguistic hedges are not utilized in the similarity calculation within fuzzy semantic similarity measures and are ignored. This paper describes a study, which aims to capture human perceptions for linguistic hedges typically used in natural language. Twelve linguistic hedges used within natural language are selected and an experiment is conducted to capture human perceptions of the impact of hedges on fuzzy category words. A dataset of hedge sentence pairs is created and rated in terms of similarity by human participants. Excellent inter-rater correlations and inter-class correlations are established between the average human ratings and an established fuzzy semantic similarity measure.
The annotation guidelines for Universal Dependencies (UD) stipulate that the basic units of dependency annotation are syntactic words, but it is not clear what are syntactic words in Japanese. Departing from the long tradition of using phrasal units called bunsetsu for dependency parsing, the current UD Japanese treebanks adopt the Short Unit Words. However, we argue that they are not syntactic word as specified by the annotation guidelines. Although we find non-mainstream attempts to linguistically define Japanese words, such definitions have never been applied to corpus annotation. We discuss the costs and benefits of adopting the rather unfamiliar criteria.
We investigate the problem of parsing conversational data of morphologically-rich languages such as Hindi where argument scrambling occurs frequently. We evaluate a state-of-the-art non-linear transition-based parsing system on a new dataset containing 506 dependency trees for sentences from Bollywood (Hindi) movie scripts and Twitter posts of Hindi monolingual speakers. We show that a dependency parser trained on a newswire treebank is strongly biased towards the canonical structures and degrades when applied to conversational data. Inspired by Transformational Generative Grammar, we mitigate the sampling bias by generating all theoretically possible alternative word orders of a clause from the existing (kernel) structures in the treebank. Training our parser on canonical and transformed structures improves performance on conversational data by around 9% LAS over the baseline newswire parser.
This contribution presents the work in progress to convert the Ancient Greek and Latin Dependency Treebank (AGLDT) into standoff annotation using PAULA XML. With an increasing number of annotations of any kind, it becomes more and more urgent that annotations related to the same texts be added standoff. Standoff annotation consists in adding any kind of annotation in separate documents, which are ultimately linked to a main text, the so-called "base text," which is meant to be unchangeable. References occur via a graph-based system of IDs, which allows an annotation layer (contained in a separate file) to be linked to another annotation layer (contained in another separate file). All the annotations/files create a labeled directed acyclic graph, whose root is represented by the base text. Standoff annotation enables easy interoperability and extension, in that single annotation layers can reference other layers of annotation independently, thus overcoming the problem of conflicting hierarchies. Moreover, standoff annotation also allows addition of different annotations of the same kind to the same text (e.g., two different interpretations of the POS tag for a given token). In the present contribution, I show how the annotations of the AGLDT can become standoff using PAULA XML, which is an open access format following the LAF principles. More precisely, I show the case study of Caesar's De Bello Civili. I detail the PAULA XML files created for its tokenization and sentence split, which are preliminary required to add morphosyntactic annotation.
In this paper, we present the current version of two different treebanks, the re-annotation of the Turkish PUD Treebank and the first annotation of the Turkish National Corpus Universal Dependency (henceforth TNC-UD). The annotation of both treebanks, the Turkish PUD Treebank and TNC-UD, was carried out based on the decisions concerning linguistic adequacy of re-annotation of the Turkish IMST-UD Treebank (Türk et. al., forthcoming). Both of the treebanks were annotated with the same annotation process and morphological and syntactic analyses. The TNC-UD is planned to have 10,000 sentences. In this paper, we will present the first 500 sentences along with the annotation PUD Treebank. Moreover, this paper also offers the parsing results of a graph-based neural parser on the previous and re-annotated PUD, as well as the TNC-UD. In light of the comparisons, even though we observe a slight decrease in the attachment scores of the Turkish PUD treebank, we demonstrate that the annotation of the TNC-UD improves the parsing accuracy of Turkish. In addition to the treebanks, we have also constructed a custom annotation software with advanced filtering and morphological editing options. Both the treebanks, including a full edit-history and the annotation guidelines, and the custom software are publicly available under an open license online.
Classical Chinese is an isolating language without notational inflection, and its texts are continuous strings of Chinese characters without spaces or punctuations between words or sentences. In order to apply Universal Dependencies for classical Chinese, we need several “not-universal” treatments and enhancements. In this paper such treatments and enhancements are revealed.
This paper presents a new schema to annotate Chinese Treebanks on the character level. The original Universal Dependencies (UD) and Surface-Syntactic Universal Dependencies (SUD) projects provide token-level resources with rich morphosyntactic language details. However, without any commonly accepted word definition for Chinese, the dependency parsing always faces the dilemma of word segmentation. Therefore we present a character-level annotation schema integrated into the existing Universal Dependencies schema as an extension.
PURPOSE: Corneal confocal microscopy (CCM) is an imaging method to detect loss of nerve fibers in the cornea. The impact of image quality on the CCM parameters has not been investigated. We developed a quality index (QI) with 3 stages for CCM images and compared the influence of the image quality on the quantification of corneal nerve parameters using 2 modes of analysis in healthy volunteers and patients with known peripheral neuropathy. METHODS: Images of 75 participants were a posteriori analyzed, including 25 each in 3 image quality groups (QI 1-QI 3). Corneal nerve fiber length (CNFL) was analyzed using automated and semiautomated software, and corneal nerve fiber density and corneal nerve branch density were quantified using automated image analysis. Three masked raters assessed CCM image quality (QI) independently and categorized images into groups QI 1-QI 3. In addition, statistical analysis was used to compare interrater reliability. Analysis of variance was used for analysis between the groups. Interrater reliability analysis between the image ratings was performed by calculating Fleiss' kappa and its 95% confidence interval. RESULTS: CNFL, corneal nerve fiber density, and corneal nerve branch density increased significantly with QI (P < 0.001, all post hoc tests P < 0.05). CNFL was higher using semiautomated compared with automated nerve analysis, independent of QI. Fleiss kappa coefficient for interrater reliability of QI was 0.72. CONCLUSIONS: The quantification of corneal nerve parameters depends on image quality, and poorer quality images are associated with lower values for corneal nerve parameters. We propose the QI as a tool to reduce variability in quantification of corneal nerve parameters.
In this paper we present the ConlluEditor annotation tool for manual annotation of files in CoNLL-U format, such as Universal Dependencies treebanks. Apart from providing a graphical editor for basic and enhanced dependencies, multi-token words, it also runs validation scripts to find potential errors. ConlluEditor uses a client-server architecture. It is freely-available under the 3-Clause BSD License.
In spite of the current availability of large collections of treebanks that can be used and queried from one common place on the web, we are still far from achieving a real interconnection, both between treebanks themselves and with other (kinds of) linguistic resources. However, making resources interoperable is a crucial requirement to maximize the contribution of each single resource, as well as to account for the linguistic complexity of the texts provided by (annotated) corpora and particularly by treebanks. This paper describes how dependency treebanks are interlinked in a Knowledge Base of linguistic resources for Latin based on Linked Open Data practices and standards. The Knowledge base is built to make linguistic resources interact by integrating all types of annotation applied to a particular word/text into a common representation.
This contribution is about the annotation of sentences with verbal predicate ellipsis in the Estonian Universal Dependencies (UD) treebank. The main aim of the UD initiative is to develop a cross-linguistically consistent treebank annotation scheme and build a multilingual treebank collection. There are more than 70 treebanks in over 100 languages in UD treebank collection version 2.2. However, the UD annotation scheme is constantly improved and amended and so the annotation of the treebanks is also changing from version to version.
This paper describes the application of annotation engineering techniques for the construction of a corpus for Role and Reference Grammar (RRG). RRG is a semantics-oriented formalism for natural language syntax popular in comparative linguistics and linguistic typology, and predominantly applied for the description of non-European languages which are less-resourced in terms of natural language processing. Because of its cross-linguistic applicability and its conjoint treatment of syntax and semantics, RRG also represents a promising framework for research challenges within natural language processing. At the moment, however, these have not been explored as no RRG corpus data is publicly available. While RRG annotations cannot be easily derived from any single treebank in existence, we suggest that they can be reliably inferred from the intersection of syntactic and semantic annotations as represented by, for example, the Universal Dependencies (UD) and PropBank (PB), and we demonstrate this for the English Web Treebank, a 250,000 token corpus of various genres of English internet text. The resulting corpus is a gold corpus for future experiments in natural language processing in the sense that it is built on existing annotations which have been created manually. A technical challenge in this context is to align UD and PB annotations, to integrate them in a coherent manner, and to distribute and to combine their information on RRG constituent and operator projections. For this purpose, we describe a framework for flexible and scalable annotation engineering based on flexible, unconstrained graph transformations of sentence graphs by means of SPARQL Update.
= 175, 20-79 years) revealed (a) a memory-experience gap for negative affect (more intense retrospective ratings than aggregated momentary ratings) that is attenuated with advancing age; (b) only a small memory-experience gap for positive affect for very old adults (66-79 years), but not younger adults; (c) relatively high convergence of aggregated momentary ratings and retrospective ratings despite (d) small biases of retrospective negative affect ratings toward peak and most recent negative affect. Findings suggest that both rating types can discriminate "good days" from "bad days" and provide overlapping but not necessarily exchangeable information. (PsycInfo Database Record (c) 2020 APA, all rights reserved).
Abstract This paper presents the Slovene Training Corpus ssj500k 2.2, which has been annotated on the levels of tokenization, sentence segmentation, part-of-speech tagging, lemmatization, syntactic dependencies, named entities, verbal multi-word expressions, and semantic role labeling. It describes the individual layers of annotation and shows the scope of using the training corpus in the production of various lexicons, such as the lexicon of multi-word units and the valency lexicon of modern Slovene. It concludes by presenting our future work, i.e. the annotation of multi-word expressions based on the Slovene Lexical Database.
Over the last 20 years, the development of a wide range of treebanks that track the evolution of languages’ syntactic patterns through time has revolutionized the field of historical syntax. The range of treebanks now available facilitates research into the long histories of many of the major Indo-European languages. Although the field's essentially corpus-based methodology has not changed, the quantity of data now available and the ease and precision with which those data can be extracted have created new opportunities. For example, with a treebank it is possible to extract all examples of surface strings associated only with abstract structures (e.g., relative clauses, extraposition), to investigate predictions made by syntactic analyses, to search for rare constructions, and to extract enough data to support sophisticated statistical analyses. Crucially, treebanks make verification and replicability of results possible.
The battle against HIV is one of the objectives in our century. Along these lines, among the HIV-contaminated patients, one of the most perilous and remarkable with its entanglements is those with lung pathologies. Concurring clinical arranging of the illness, such patients may introduce Tuberculosis, Pneumocystis jirovecii, Cytomegaloviruses, Candidiasis, Toxoplasmosis and so on. The exploration by logical examination foundation of lung illness was done among the inpatient people in measure of 48.37 (77%) of them were given tuberculosis and 11 (23%) with Interstitial Lung Disease (ILD). Studies were introduced on HIV-positive patients who were isolated by the randomization methods. Among 37 patients with tuberculosis, 29 (78%) had AFB (corrosive quick bacillius) with Gexpert, HAIN strategies, 6 (22%) were analyzed by imaging techniques (HRCT, chest X-beam) and serum ADA level. As per past investigations, there were no relationships between's serum ADA level rises at HIV-positive patients (p esteem 0.05). Among 11 patients gave ILD Pneumocystis jirovecii were distinguished at 5 (45%), 3 (27.5%) were given every day mortality, 3 took a Co-Trimaxozole treatment analyzed by imaging strategies. Clinical viability was endorsed by the nearness of pneumocystis starting point. At the second phase of the examination was discovered a relationship between's various Cd4 cell check and imaging rating. In this way, among absolute number of 119 HIV-positive patients, 38 (32%) had penetration zones, 53 (44%) had an obliteration, 20 (17%) dispersal, 8 (7%) mediastinal lymphadenopathy. Measurement results p esteem 0.000424, hence there is immediate relationship. There are heap aspiratory conditions related with HIV, extending from intense contaminations to constant noncommunicable infections. The study of disease transmission of these illnesses has changed essentially in the time of far reaching antiretroviral treatment. Assessment of the HIV-tainted patient includes evaluation of the seriousness of ailment and an intensive yet effective quest for authoritative conclusion, which may include various etiologies at the same time. Significant pieces of information to an analysis incorporate clinical and social history, segment subtleties, for example, travel and geology of habitation, substance use, sexual practices, and domiciliary and detainment status. CD4 cell tally is a colossally valuable proportion of resistant capacity and hazard for HIV-related infections, and assists slender with bringing down the differential. Cautious history of current side effects and physical assessment with specific thoughtfulness regarding extrapulmonary signs are vital early advances. Numerous adjunctive research facility studies can recommend or preclude specific findings. Pneumonic capacity testing (PFT) may help in portrayal of a few incessant noninfectious sicknesses quickened by HIV. Chest radiograph and registered tomography (CT) check take into account arrangement of sicknesses by pathognomonic imaging designs, albeit numerous irresistible conditions present atypically, especially with lower CD4 tallies. At last, conclusive finding with sputum, bronchoscopy with bronchoalveolar lavage, or lung tissue is regularly required. It is of most extreme significance to keep up a serious extent of doubt for HIV in any case undiscovered patients, as the principal introduction of HIV might be through an intense pneumonic disease. The assessment of respiratory side effects in HIV-contaminated patients can be trying for various reasons. Respiratory manifestations are a continuous objection among HIV-contaminated people and might be brought about by a wide range of ailments. The range of pneumonic sicknesses in HIV-tainted patients incorporates both HIV-related and non-HIV-related conditions. The HIV-related aspiratory conditions incorporate both entrepreneurial contaminations (OIs) and neoplasms. The OIs include bacterial, mycobacterial, contagious, viral, and parasitic pathogens. Every one of these OIs and neoplasms has a trademark clinical and radiographic introduction. Nonetheless, there can be impressive variety and cover in these introductions. Thusly, no group of stars of manifestations, physical assessment discoveries, research facility anomalies, and chest radiographic discoveries is pathognomonic or explicit for a specific malady. Thus, a complete microbiologic or pathologic determination is desirable over empiric treatment at whatever point conceivable. Symptomatic tests incorporate societies from sputum and blood and from respiratory examples acquired by obtrusive systems, for example, bronchoscopy, thoracentesis, figured tomography (CT)- guided transthoracic needle goal, thoracoscopy, mediastinoscopy, and open-lung biopsy. This section portrays the recurrence of respiratory side effects, the range of pneumonic diseases that can influence HIV-tainted patients, and an indicative way to deal with the assessment of respiratory manifestations in HIV-contaminated patients, featuring certain parts of the clinical introduction that might be valuable in separating the most widely recognized OIs and neoplasms. The trademark chest radiographic introductions of the most well-known OIs and neoplasms are portrayed and blueprints of 3 case situations are introduced to outline differential judgments and significant analytic and restorative choices for an assortment of clinical and radiographic introductions. For subtleties on explicit indicative tests and treatment regimens for every one of the OIs and neoplasms, see explicit parts inside the HIV InSite Knowledge Base.
Despite advances in dependency parsing, languages with small treebanks still present challenges.
We show that sampling latent variables multiple times at a gradient step helps in improving a variational autoencoder and propose a simple and effective method to better exploit these latent variables through hidden state averaging. Consistent gains in performance on two different datasets, Penn Treebank and Yahoo, indicate the generalizability of our method. 1
Cognate vocabulary is known to have a facilitating effect on foreign language (L2) lexical development (de Groot and Keijzer, 2000; Elgort, 2013). Because of their cross-lingual semiotic transparency, cognates are known to be easier to comprehend and learn. As a result, cognate status has been considered an important feature when modeling L2 vocabulary learning (Willis and Ohashi, 2012) or when assessing L2 lexical readability (Beinborn et al., 2014). Although the latter readability-focused user study has shown a positive effect of cognates on decontextualized word comprehension, not many studies seem to have focused on how cognate vocabulary is distributed in reading texts of different L2 levels, such as reading materials found in textbooks graded along the CEFR (Common European Framework of Reference) scale (Council of Europe, 2001). Our aim is therefore to examine whether the presupposed increasing difficulty of the lexical stock attested in such texts is somehow related to cognate density. To this end, we will focus on French and Dutch L2 and will use two lexical databases, viz. FLELex (Francois et al., 2014) and NT2Lex (Tack et al., 2018), respectively. These resources have been compiled from a corpus of L2 reading materials targeted towards a specific CEFR level, including expert-written texts found in textbooks or readers. The lexicons thus describe word frequency distributions observed along the CEFR scale and therefore inform us about the lexical stock that should be understood a priori at a given level. In these CEFR-graded word distributions, cognate vocabulary in Dutch and French will be automatically identified, drawing on recent machine translation methods (Beinborn et al., 2013; Mitkov et al., 2007). As a parallel reference dataset, we will use the Dutch-French alignments of the Dutch Parallel Corpus (Paulussen et al., 2006).
Neural networks have achieved great success in many difficult learning tasks like image classification, speech recognition and natural language processing. However, neural architectures are hard to design, which requires lots of knowledge and time of human experts. Therefore, there has been a growing interest in automating the process of designing neural architectures. Though these searched architectures have achieved competitive performance on various tasks, the efficiency of NAS still needs to be improved. Moreover, current neural architecture search approach disregards the dependency between a node and its predecessors and successors. <br/>This thesis builds upon BayesNAS which employs the classic Bayesian learning method to search for CNN architectures, and extends it to the problem of neural architecture search for recurrent architectures. Hierarchical sparse priors are used to model the architecture parameters to alleviate the dependency issue. Since the update of posterior variance is based on Laplace approximation, an efficient method to compute the Hessian of recurrent layer is proposed. We can find candidated architectures after training the over-parameterized network for only one epoch. Our experiments on Penn Treebank and WikiText-2 show that competitive architectures can be found in 0.3 GPU days using a single GPU for language modeling task. We find that our algorithm is more efficient than state-of-the-art.<br/>