Universal Dependencies Research Articles

Treebanking user-generated content: a UD based overview of guidelines, corpora and unified recommendations

This article presents a discussion on the main linguistic phenomena which cause difficulties in the analysis of user-generated texts found on the web and in social media, and proposes a set of annotation guidelines for their treatment within the Universal Dependencies (UD) framework of syntactic analysis. Given on the one hand the increasing number of treebanks featuring user-generated content, and its somewhat inconsistent treatment in these resources on the other, the aim of this article is twofold: (1) to provide a condensed, though comprehensive, overview of such treebanks—based on available literature—along with their main features and a comparative analysis of their annotation criteria, and (2) to propose a set of tentative UD-based annotation guidelines, to promote consistent treatment of the particular phenomena found in these types of texts. The overarching goal of this article is to provide a common framework for researchers interested in developing similar resources in UD, thus promoting cross-linguistic consistency, which is a principle that has always been central to the spirit of UD.

Two languages, one treebank: building a Turkish–German code-switching treebank and its challenges

This paper presents the SAGT Turkish–German code-switching treebank, and observations and annotation challenges we encountered during its development. The treebank consists of transcriptions of bilingual conversations annotated with several layers: language IDs, lemmas, POS tags, morphological features, and dependency relations. The annotations follow the Universal Dependencies annotation scheme and the conventions used in monolingual treebanks as much as possible. We present and discuss a number of issues that arise because of the need for consistent multilingual annotation within a single treebank, as well as the informal language, which is where code-switching is observed most. Besides proposing solutions to these issues, we present some observations about code-switching phenomena that are only possible to observe in a data set with rich linguistic annotation. The treebank was annotated with a focus on quality of annotations through an iterative process of detecting and correcting annotation errors. We also present quantitative measures for indication of annotation quality. The code-switching treebank created in this study is released to the public through Universal Dependencies repositories.

Joint learning of morphology and syntax with cross-level contextual information flow

AbstractWe propose an integrated deep learning model for morphological segmentation, morpheme tagging, part-of-speech (POS) tagging, and syntactic parsing onto dependencies, using cross-level contextual information flow for every word, from segments to dependencies, with an attention mechanism at horizontal flow. Our model extends the work of Nguyen and Verspoor ((2018). Proceedings of the CoNLL Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies. The Association for Computational Linguistics, pp. 81–91.) on joint POS tagging and dependency parsing to also include morphological segmentation and morphological tagging. We report our results on several languages. Primary focus is agglutination in morphology, in particular Turkish morphology, for which we demonstrate improved performance compared to models trained for individual tasks. Being one of the earlier efforts in joint modeling of syntax and morphology along with dependencies, we discuss prospective guidelines for future comparison.

Discovering Lexical Similarity Using Articulatory Feature-Based Phonetic Edit Distance

Lexical Similarity (LS) between two languages uncovers many interesting linguistic insights such as genetic relationship, mutual intelligibility, and the usage of one's vocabulary into other. There are various methods through which LS is evaluated. In the same regard, this paper presents a method of Phonetic Edit Distance (PED) that uses a soft comparison of letters using the articulatory features associated with them. The system converts the words into the corresponding International Phonetic Alphabet (IPA), followed by the conversion of IPA into its set of articulatory features. Later, the lists of the set of articulatory features are compared using the proposed method. As an example, PED gives edit distance of German word vater and Persian word pidar as 0.82; and similarly, Hebrew word shalom and Arabic word salaam as 0.93, whereas for a juxtapose comparison, their IPA based edit distances are 4 and 2 respectively. Experiments are performed with six languages (Arabic, Hindi, Marathi, Persian, Sanskrit, and Urdu). In this regard, we extracted part of speech wise word-lists from the Universal Dependency corpora and evaluated the LS for every pair of language. Thus, with the proposed approach, we find the genetic affinity, similarity, and borrowing/loan-words despite having script differences and sound variation phenomena among these languages.

Quantifying syntax similarity with a polynomial representation of dependency trees

We introduce a graph polynomial that distinguishes tree structures to represent dependency grammar and a measure based on the polynomial representation to quantify syntax similarity. The polynomial encodes accurate and comprehensive information about the dependency structure and dependency relations of words in a sentence, which enables in-depth analysis of dependency trees with data analysis tools. We apply the polynomial-based methods to analyze sentences in the ParallelUniversal Dependencies treebanks. Specifically, we compare the syntax of sentences and their translations in different languages, and we perform a syntactic typology study of available languages in the Parallel Universal Dependencies treebanks. We also demonstrate and discuss the potential of the methods in measuring syntax diversity of corpora.

Method for the influence design assessment of a wheeled agricultural vehicle on the soil

The impact of a wheeled agricultural vehicle on the deformable support surface determines the vehicle's ability to move, as well as soil compaction, which is not desirable in agriculture. The agricultural machine must not cause more pressure on the ground than is permissible. Therefore, in the tasks of design numerical modelling of the agricultural vehicle movement or trailer, it is required to calculate the specified parameter. It is impossible to calculate without knowledge of the geometric characteristics of the contact spot associated with the normal deformation of the tire under normal load. To calculate these characteristics, it is necessary to have universal dependencies for determining the normal stiffness of the tire. These are available for tires of various purposes. The elastic properties of ultra-low pressure tires are insufficiently studied. Experimental studies of the elastic properties of these tires have been carried out with the authors participation. However, there are currently no dependencies to describe them. This does not provide the possibility of a correct design calculation of the influence of such tires on the soil. The purpose of the work: to develop a universal method for calculating the influence on the soil of agricultural vehicle. A universal method for calculating the impact of a wheeled agricultural machine on the ground has been developed. Universal design-experimental dependence for determining the normal stiffness of ultra-low pressure tires is obtained. It takes into account tire pressure, normal load under specific conditions and geometric characteristics.

Syntactic Factors Associated With Performance of Dependency Parsers Using Stack-Pointer Network and Graph Attention Networks Between English and Korean

A Stack-Pointer Network (StackPtr) parser is a pointer network with an internal stack on the decoder. Several studies use the StackPtr as the backbone of a dependency parser because it can traverse a parse tree depth-first without backtracking and can handle high-order parsing information easily thanks to the internal stack. The parser can use information from previously derived subtrees stored in the internal stack upon selecting a child node. In this work, we introduce a new StackPtr parser with Graph Attention Networks (GATs) that can encode a previously derived subtree. We evaluated our proposed parser on the Sejong and Everyone’s corpora for Korean and on the Penn Treebank and Universal Dependency corpora for English. In addition, we analyzed and compared our proposed parser with other variants of the StackPtr parser, examining the syntactic information that each parser can reference at every decoding step. We found that Korean parse trees tend to have more consecutive immediate single-child nodes than English parse trees. The proposed StackPtr parser with GATs performed best on almost all metrics for Korean because it can utilize more context to analyze these parse trees by grasping Korean syntactic factors than any other variants. However, for English, no particular variant of the StackPtr parser outperforms the others.

Sputtering, Cluster Primary Ions and Static SIMS

Sputtering using cluster primary ion beams is very important for the future development of static SIMS and the SIMS depth profiling of organic layers. However, different results from different laboratories may be confusing. Analytical models have an important function for enabling the prediction of behaviour for practical analysis. Sigmund's model for sputtering, often used in surface analysis, is helpful and accurate in the linear cascade regime. However, for cluster sputtering this is no longer the case and spike effects need evaluation. Evidence will be presented of the spike model validity for clusters of up to more than 10 atoms over 3 orders of magnitude in sputtering yield. Using data from one primary ion, extremely good descriptions of measurements reported with other primary ions can then be achieved. This theory is then used to evaluate the molecular ion yield behaviour of interest in the static SIMS of organics. This leads to universal dependencies for the de-protonated molecular ion yields, relating all primary ions, both single atom and cluster, which are illustrated by experimental data over 5 decades of emission intensity. This formulation permits the prediction of the (M-H)- secondary ion yield for different, or new, primary ion sources. It is shown how further gains are predicted. For analysing materials, raising the molecular secondary ion yield is extremely helpful but it is the ratio of this yield to the disappearance cross-section (the efficiency) that is critical. The relation of the damage and disappearance cross sections is formulated. Data are evaluated and a description is given to show how these cross sections are related and to provide a further universal relation for the efficiency/yield dependence of all cluster ions.

Tri-training for Dependency Parsing Domain Adaptation

In recent years, the research on dependency parsing focuses on improving the accuracy of the domain-specific (in-domain) test datasets and has made remarkable progress. However, there are innumerable scenarios in the real world that are not covered by the dataset, namely, the out-of-domain dataset. As a result, parsers that perform well on the in-domain data usually suffer from significant performance degradation on the out-of-domain data. Therefore, to adapt the existing in-domain parsers with high performance to a new domain scenario, cross-domain transfer learning methods are essential to solve the domain problem in parsing. This paper examines two scenarios for cross-domain transfer learning: semi-supervised and unsupervised cross-domain transfer learning. Specifically, we adopt a pre-trained language model BERT for training on the source domain (in-domain) data at the subword level and introduce self-training methods varied from tri-training for these two scenarios. The evaluation results on the NLPCC-2019 shared task and universal dependency parsing task indicate the effectiveness of the adopted approaches on cross-domain transfer learning and show the potential of self-learning to cross-lingual transfer learning.

An HMM-Based PoS Tagger for Old Church Slavonic

Abstract We present a hybrid HMM-based PoS tagger for Old Church Slavonic. The training corpus is a portion of one text, Codex Marianus (40k) annotated with the Universal Dependencies UPOS tags in the UD-PROIEL treebank. We perform a number of experiments in within-domain and out-of-domain settings, in which the remaining part of Codex Marianus serves as a within-domain test set, and Kiev Folia is used as an out-of-domain test set. Analysing by-PoS-class precision and sensitivity in each run, we combine a simple context-free n-gram-based approach and Hidden Markov method (HMM), and added linguistic rules for specific cases such as punctuation and digits. While the model achieves a rather non-impressive accuracy of 81% in in-domain settings, we observe an accuracy of 51% in out-of-domain evaluation, which is comparable to the results of large neural architectures based on pre-trained contextual embeddings.

Resources for Turkish dependency parsing: introducing the BOUN Treebank and the BoAT annotation tool

In this paper, we introduce the resources that we developed for Turkish dependency parsing, which include a novel manually annotated treebank (BOUN Treebank), along with the guidelines we adopted, and a new annotation tool (BoAT). The manual annotation process that we employed was shaped and implemented by a team of four linguists and five Natural Language Processing (NLP) specialists. Decisions regarding the annotation of the BOUN Treebank were made in line with the Universal Dependencies (UD) framework as well as our recent efforts for unifying the Turkish UD treebanks through manual re-annotation. To the best of our knowledge, the BOUN Treebank is the largest Turkish UD treebank. It contains a total of 9761 sentences from various topics including biographical texts, national newspapers, instructional texts, popular culture articles, and essays. In addition, we report the parsing results of a state-of-the-art dependency parser obtained over the BOUN Treebank as well as two other treebanks in Turkish. Our results demonstrate that the unification of the Turkish annotation scheme and the introduction of a more comprehensive treebank lead to improved performance with regards to dependency parsing.

Semantic Roles or Syntactic Functions: The Effects of Annotation Scheme on the Results of Dependency Measures

AbstractThe annotation scheme of dependency treebanks might have an impact on the results of linguistic analysis, thus leading to different interpretations of linguistic phenomena. This study compares the results of two widely used dependency measures, i.e., dependency direction and dependency distance, based on 18 parallel Universal Dependencies (UD) annotated treebanks and 18 corresponding Surface‐Syntactic Universal Dependencies (SUD) annotated treebanks. The results show that (1) Based on the semantic UD and syntactic SUD, dependency relations between function words and content words share the opposite dependency directions but similar dependency distances; (2) Annotation scheme has a significant impact on dependency direction, though the effect size is small. We find that the proportions of head‐final dependencies based on the syntactic SUD can better group language families than those based on semantic UD; (3) Annotation scheme also affects dependency distance significantly, though its effect size is small. Mean dependency distances (MDDs) based on UD are always higher than those based on SUD. However, the MDDs based on both annotation schemes are within a certain threshold, which shows that the linguistic universal of dependency distance minimization is independent of annotation schemes.

Integration of morphological features and contextual weightage using monotonic chunk attention for part of speech tagging

Part-of-Speech (POS) tagging is a fundamental sequence labeling problem in Natural Language Processing. Recent deep learning sequential models combine the forward and backward word informatio for POS tagging. The information of contextual words to the current word play a vital role in capturing the non-continuous relationship. We have proposed Monotonic chunk-wise attention with CNN-GRU-Softmax (MCCGS), a deep learning architecture that adheres to these essential information. This architecture consists of Input Encoder (IE), encodes word and character-level, Contextual Encoder (CE), assigns the weightage to adjacent word and Disambiguator (D), which resolves intra-label dependencies as core components. Moreover, different morphological features have been integrated into the core components of MCCGS architecture as MCCGS-IE, MCCGS-CE and MCCGS-D. The MCCGS architecture is validated on the 21 languages from Universal Dependency (UD) treebank. The state-of-the-art models, Type constraints, Retrofitting, Distant Supervision from Disparate Sources and Position-aware Self Attention, MCCGS and its variants such as MCCGS-IE, MCCGS-CE and MCCGS-D are obtained mean accuracy 83.65%, 81.29%, 84.10%, 90.18%, 90.40%, 91.40%, 90.90%, 92.30%, respectively. The proposed model architecture provides state-of-the-art accuracy on the low resource languages as Marathi (93.58%), Tamil (87.50%), Telugu (96.69%) and Sanskrit (97.28%) from UD treebank and Hindi (95.64%) and Urdu (87.47%) from Hindi-Urdu multi-representational treebank.

Thermal Conductivity of Cellulose Fibers in Different Size Scales and Densities.

Considering the growing use of cellulose in various applications, knowledge and understanding of its physical properties become increasingly important. Thermal conductivity is a key property, but its variation with porosity and density is unknown, and it is not known if such a variation is affected by fiber size and temperature. Here, we determine the relationships by measurements of the thermal conductivity of cellulose fibers (CFs) and cellulose nanofibers (CNFs) derived from commercial birch pulp as a function of pressure and temperature. The results show that the thermal conductivity varies relatively weakly with density (ρsample = 1340–1560 kg m–3) and that its temperature dependence is independent of density, porosity, and fiber size for temperatures in the range 80–380 K. The universal temperature and density dependencies of the thermal conductivity of a random network of CNFs are described by a third-order polynomial function (SI-units): κCNF = (0.0787 + 2.73 × 10–3·T – 7.6749 × 10–6·T2 + 8.4637 × 10–9·T3)·(ρsample/ρ0)2, where ρ0 = 1340 kg m–3 and κCF = 1.065·κCNF. Despite a relatively high degree of crystallinity, both CF and CNF samples show amorphous-like thermal conductivity, that is, it increases with increasing temperature. This appears to be due to the nano-sized elementary fibrils of cellulose, which explains that the thermal conductivity of CNFs and CFs shows identical behavior and differs by only ca. 6%. The nano-sized fibrils effectively limit the phonon mean free path to a few nanometers for heat conduction across fibers, and it is only significantly longer for highly directed heat conduction along fibers. This feature of cellulose makes it easier to apply in applications that require low thermal conductivity combined with high strength; the weak density dependence of the thermal conductivity is a particularly useful property when the material is subjected to high loads. The results for thermal conductivity also suggest that the crystalline structures of cellulose remain stable up to at least 0.7 GPa.

Joint Universal Syntactic and Semantic Parsing

While numerous attempts have been made to jointly parse syntax and semantics, high performance in one domain typically comes at the price of performance in the other. This trade-off contradicts the large body of research focusing on the rich interactions at the syntax–semantics interface. We explore multiple model architectures that allow us to exploit the rich syntactic and semantic annotations contained in the Universal Decompositional Semantics (UDS) dataset, jointly parsing Universal Dependencies and UDS to obtain state-of-the-art results in both formalisms. We analyze the behavior of a joint model of syntax and semantics, finding patterns supported by linguistic theory at the syntax–semantics interface. We then investigate to what degree joint modeling generalizes to a multilingual setting, where we find similar trends across 8 languages.

Cross-lingual transfer learning for relation extraction using Universal Dependencies

This paper focuses on the task of cross-language relation extraction, which aims to identify the semantic relations holding between entities in the text. The goal of the task is to train classifiers for low-resource languages by means of the annotated data from high-resource languages. Related methods usually employ parallel data or Machine Translator (MT) to project annotated data from a source to a target language. However, the availability and the quality of parallel data and MT are big challenges for low-resource languages. In this paper, a novel transfer learning method is presented for this task. The key idea is to utilize a tree-based representation of data, which is highly informative for classifying semantic relations, and also shared among different languages. All the training and test data are shown using this representation. We propose to use the Universal Dependency (UD) parsing, which is a language-agnostic formalism for representation of syntactic structures. Equipping UD parse trees with multi-lingual word embeddings makes an ideal representation for the cross-language relation extraction task. We propose two deep networks to use this representation. The first one utilizes the Shortest Dependency Path of UD trees, while the second employs the UD-based positional embeddings. Experiments are performed using SemEval 2010-task 8 training data, whereas French and Farsi are the test languages. The results show 63.9% and 56.2% F1 scores, for French and Farsi test data, respectively, which are 14.4% and 17.9% higher than the baseline. This work can be considered a simple yet powerful baseline for further investigation into the cross-language tasks.

Antitumor efficacy of XPO1 inhibitor Selinexor in KRAS-mutant lung adenocarcinoma patient-derived xenografts

Gain-of-function Kirsten rat sarcoma viral oncogene homolog (KRAS) mutations occur in 25% of lung adenocarcinomas, and these tumors are challenging to treat. Some preclinical work, largely based on cell lines, suggested KRASmut lung cancers are especially dependent on the nuclear export protein exportin-1 (XPO1), while other work supports XPO1 being a broader cancer dependency. To investigate the sensitivity of KRASmut lung cancers to XPO1 inhibition in models that more closely match clinical tumors, we treated 10 independently established lung cancer patient-derived tumor xenografts (PDXs) with the clinical XPO1 inhibitor, Selinexor. Monotherapy with Selinexor reduced tumor growth in all KRASmut PDXs, which included 4 different codon mutations, and was more effective than the clinical MEK1/2 inhibitor, Trametinib. Selinexor was equally effective in KRASG12C and KRASG12D tumors, with TP53 mutations being a biomarker for a weaker drug response. By mining genome-wide dropout datasets, we identified XPO1 as a universal cancer cell dependency and confirmed this functionally in two KRASWT PDX models harboring kinase drivers. However, targeted kinase inhibitors were more effective than Selinexor in these models. Our findings support continued investigation of XPO1 inhibitors in KRASmut lung adenocarcinoma, regardless of the codon alteration.

Cross-Linguistic Trade-Offs and Causal Relationships Between Cues to Grammatical Subject and Object, and the Problem of Efficiency-Related Explanations.

Cross-linguistic studies focus on inverse correlations (trade-offs) between linguistic variables that reflect different cues to linguistic meanings. For example, if a language has no case marking, it is likely to rely on word order as a cue for identification of grammatical roles. Such inverse correlations are interpreted as manifestations of language users’ tendency to use language efficiently. The present study argues that this interpretation is problematic. Linguistic variables, such as the presence of case, or flexibility of word order, are aggregate properties, which do not represent the use of linguistic cues in context directly. Still, such variables can be useful for circumscribing the potential role of communicative efficiency in language evolution, if we move from cross-linguistic trade-offs to multivariate causal networks. This idea is illustrated by a case study of linguistic variables related to four types of Subject and Object cues: case marking, rigid word order of Subject and Object, tight semantics and verb-medial order. The variables are obtained from online language corpora in thirty languages, annotated with the Universal Dependencies. The causal model suggests that the relationships between the variables can be explained predominantly by sociolinguistic factors, leaving little space for a potential impact of efficient linguistic behavior.

CO<sub>2</sub>-equivalence metrics for surface albedo change based on the radiative forcing concept: a critical review

Abstract. Management of Earth's surface albedo is increasingly viewed as an important climate change mitigation strategy both on (Seneviratne et al., 2018) and off (Field et al., 2018; Kravitz et al., 2018) the land. Assessing the impact of a surface albedo change involves employing a measure like radiative forcing (RF) which can be challenging to digest for decision-makers who deal in the currency of CO2-equivalent emissions. As a result, many researchers express albedo change (Δα) RFs in terms of their CO2-equivalent effects, despite the lack of a standard method for doing so, such as there is for emissions of well-mixed greenhouse gases (WMGHGs; e.g., IPCC AR5, Myhre et al., 2013). A major challenge for converting Δα RFs into their CO2-equivalent effects in a manner consistent with current IPCC emission metric approaches stems from the lack of a universal time dependency following the perturbation (perturbation “lifetime”). Here, we review existing methodologies based on the RF concept with the goal of highlighting the context(s) in which the resulting CO2-equivalent metrics may or may not have merit. To our knowledge this is the first review dedicated entirely to the topic since the first CO2-eq. metric for Δα surfaced 20 years ago. We find that, although there are some methods that sufficiently address the time-dependency issue, none address or sufficiently account for the spatial disparity between the climate response to CO2 emissions and Δα – a major critique of Δα metrics based on the RF concept (Jones et al., 2013). We conclude that considerable research efforts are needed to build consensus surrounding the RF “efficacy” of various surface forcing types associated with Δα (e.g., crop change, forest harvest), and the degree to which these are sensitive to the spatial pattern, extent, and magnitude of the underlying surface forcings.

Energy and flux budget closure theory for passive scalar in stably stratified turbulence

The energy and flux budget (EFB) closure theory for a passive scalar (non-buoyant and non-inertial particles or gaseous admixtures) is developed for stably stratified turbulence. The physical background of the EFB turbulence closures is based on the budget equations for the turbulent kinetic and potential energies and turbulent fluxes of momentum and buoyancy as well as the turbulent flux of particles. The EFB turbulence closure is designed for stratified geophysical flows from neutral to very stable stratification, and it implies that turbulence is maintained by the velocity shear at any stratification. In a steady-state, expressions for the turbulent flux of the passive scalar and the anisotropic non-symmetric turbulent diffusion tensor are derived, and universal flux Richardson number dependencies of the components of this tensor are obtained. The diagonal component in the vertical direction of the turbulent diffusion tensor is suppressed by strong stratification, while the diagonal components in the horizontal directions are not suppressed, but they are dominant in comparison with the other components of the turbulent diffusion tensor. This implies that any initially created strongly inhomogeneous particle cloud is evolved into a thin pancake in a horizontal plane with very slow increase in its thickness in the vertical direction. The turbulent Schmidt number (the ratio of the eddy viscosity and the vertical turbulent diffusivity of the passive scalar) linearly increases with the gradient Richardson number. The physics of such a behavior is related to the buoyancy force that causes a correlation between fluctuations of the potential temperature and the particle number density. This correlation that is proportional to the product of the vertical turbulent particle flux and the vertical gradient of the mean potential temperature reduces the vertical turbulent particle flux. Considering the applications of these results to the atmospheric boundary-layer turbulence, the theoretical relationships are derived, which allows us to determine the turbulent diffusion tensor as a function of the vertical coordinate measured in the units of the local Obukhov length scale. The obtained relations are potentially useful in modeling applications of particle dispersion in the atmospheric boundary-layer turbulence and free atmosphere turbulence.