Further evidence for a parent-of-origin effect at the NOP9 locus on language-related phenotypes

Background - Specific language impairment (SLI) is a common neurodevelopmental disorder, observed in 5–10 % of children. Family and twin studies suggest a strong genetic component, but relatively few candidate genes have been reported to date. A recent genome-wide association study (GWAS) described the first statistically significant association specifically for a SLI cohort between a missense variant (rs4280164) in the NOP9 gene and language-related phenotypes under a parent-of-origin model. Replications of these findings are particularly challenging because the availability of parental DNA is required. Methods - We used two independent family-based cohorts characterised with reading- and language-related traits: a longitudinal cohort (n = 106 informative families) including children with language and reading difficulties and a nuclear family cohort (n = 264 families) selected for dyslexia. Results - We observed association with language-related measures when modelling for parent-of-origin effects at the NOP9 locus in both cohorts: minimum P = 0.001 for phonological awareness with a paternal effect in the first cohort and minimum P = 0.0004 for irregular word reading with a maternal effect in the second cohort. Allelic and parental trends were not consistent when compared to the original study. Conclusions - A parent-of-origin effect at this locus was detected in both cohorts, albeit with different trends. These findings contribute in interpreting the original GWAS report and support further investigations of the NOP9 locus and its role in language-related traits. A systematic evaluation of parent-of-origin effects in genetic association studies has the potential to reveal novel mechanisms underlying complex traits.

Multilingualism and conceptual modelling

One of the leading motivations behind the multilingual semantic web is to make resources accessible digitally in an online global multilingual context. Consequently, it is fundamental for knowledge bases to find a way to manage multilingualism and thus be equipped with those procedures for its conceptual modelling. In this context, the goal of this paper is to discuss how common-sense knowledge and cultural knowledge are modelled in a multilingual framework. More particularly, multilingualism and conceptual modelling are dealt with from the perspective of FunGramKB, a lexico-conceptual knowledge base for natural language understanding. This project argues for a clear division between the lexical and the conceptual dimensions of knowledge. Moreover, the conceptual layer is organized into three modules, which result from a strong commitment towards capturing semantic knowledge (Ontology), procedural knowledge (Cognicon) and episodic knowledge (Onomasticon). Cultural mismatches are discussed and formally represented at the three conceptual levels of FunGramKB.

Face recognition using a unified 3D morphable model

We address the problem of 3D-assisted 2D face recognition in scenarios when the input image is subject to degradations or exhibits intra-personal variations not captured by the 3D model. The proposed solution involves a novel approach to learn a subspace spanned by perturbations caused by the missing modes of variation and image degradations, using 3D face data reconstructed from 2D images rather than 3D capture. This is accomplished by modelling the difference in the texture map of the 3D aligned input and reference images. A training set of these texture maps then defines a perturbation space which can be represented using PCA bases. Assuming that the image perturbation subspace is orthogonal to the 3D face model space, then these additive components can be recovered from an unseen input image, resulting in an improved fit of the 3D face model. The linearity of the model leads to efficient fitting. Experiments show that our method achieves very competitive face recognition performance on Multi-PIE and AR databases. We also present baseline face recognition results on a new data set exhibiting combined pose and illumination variations as well as occlusion.

Modelling continuum mechanics phenomena using three dimensional unstructured meshes on massively parallel processors

The difficulties encountered in implementing large scale CM codes on multiprocessor systems are now fairly well understood. Despite the claims of shared memory architecture manufacturers to provide effective parallelizing compilers, these have not proved to be adequate for large or complex programs. Significant programmer effort is usually required to achieve reasonable parallel efficiencies on significant numbers of processors. The paradigm of Single Program Multi Data (SPMD) domain decomposition with message passing, where each processor runs the same code on a subdomain of the problem, communicating through exchange of messages, has for some time been demonstrated to provide the required level of efficiency, scalability, and portability across both shared and distributed memory systems, without the need to re-author the code into a new language or even to support differing message passing implementations. Extension of the methods into three dimensions has been enabled through the engineering of PHYSICA, a framework for supporting 3D, unstructured mesh and continuum mechanics modeling. In PHYSICA, six inspectors are used. Part of the challenge for automation of parallelization is being able to prove the equivalence of inspectors so that they can be merged into as few as possible.

Vocabulary and language model adaptation using information retrieval

International audience

Our own language: Scots verse translation and the second-generation Scottish renaissance

This thesis examines the role of Scots language verse translation in the second-generation or post-war Scottish Renaissance. The translation of European poetry into Scots was of central importance to the first-generation Scottish Renaissance of the nineteen twenties and thirties. As Margery Palmer McCulloch has shown, the wider cultural climate of Anglo-American modernism was key to MacDiarmid’s conception of the interwar Scottish Renaissance. What was the effect on second-generation poet-translators as the modernist moment passed? Are the many translations undertaken by the younger poets who emerged in the course of the nineteen forties and fifties a faithful reflection of this cultural inheritance? To what extent are they indicative of a new set of priorities and international influences? The five principal translators discussed in this thesis are Douglas Young (1913-1973), Sydney Goodsir Smith (1915-1975), Robert Garioch (1909-1981), Tom Scott (1918-1995) and William J. Tait (1918-1992). Each is the subject of a chapter, in many cases providing the first or most extensive treatment of particular translations. While the pioneering work of John Corbett, Bill Findlay and J. Derrick McClure, among other scholars, has drawn attention to the long history of literary translation into Scots, this thesis is the first extended critical work to take the verse translations of the post-MacDiarmid makars as its subject. The nature and extent of MacDiarmid’s influence is considered throughout, as are the wider discourses around language and translation in twentieth-century Scottish poetry. Critical engagement with a number of key insights from theoretical translation studies helps to situate these writers’ work in its global context. This thesis also explores the ways in which the specific context of Scots translation allows scholars to complicate or expand upon theories of translation developed in other cultural situations (notably Lawrence Venuti’s writing on domestication and foreignisation). The five writers upon whom this thesis concentrates were all highly individual, occasionally idiosyncratic personalities. Young’s polyglot ingenuity finds a foil in Garioch’s sharp, humane wit. Goodsir Smith’s romantic ironising meets its match in Scott’s radical certainty of cause. Tait’s use of the Shetlandic tongue sets him apart. Nonetheless, despite the great variety of style, form and tone shown by each of these translators, this thesis demonstrates that there are meaningful links to be made between them and that they form a unified, coherent group in the wider landscape of twentieth-century Scottish poetry. On the linguistic level, each engaged to some extent in the composition of a ‘synthetic’ or ‘plastic’ language deriving partly from literary sources, partly from the spoken language around them. On a more fundamental level, each was committed to enriching this language through translation, within which a number of key areas of interest emerge. One of the most important of these key areas is Gaelic – especially the poetry of Sorley MacLean, which Young, Garioch and Goodsir Smith all translated into Scots. This is to some extent an act of solidarity on the part of these Scots poets, acknowledging a shared history of marginalisation as well as expressing shared hopes for the future. The same is true of Goodsir Smith’s translations from a number of Eastern European poets (and Edwin Morgan’s own versions, slightly later in the century). The translation of verse drama by poets is another key theme sustained throughout the thesis, with Garioch and Young attempting to fill what they perceived as a gap in the Scots tradition through translation from other languages (another aspect of these writers’ legacy continued by Morgan). Beyond this, all of the writers discussed in this thesis translated extensively from European poetries from Ancient Greece to twentieth-century France. Their reasons for doing so were various, but a certain cosmopolitan idealism figures highly among them. So too does a desire to see Scotland interact with other European nations, thus escaping the potentially narrowing influence of post-war British culture. This thesis addresses the legacy of these writers’ translations, which, it argues, continue to exercise a perceptible influence on the course of poetry in Scotland. This work constitutes a significant contribution to a much-needed wider critical re-assessment of this pivotal period in modern Scottish writing, offering a fresh perspective on the formal and linguistic merits of these poets’ verse translations. Drawing upon frequently obscure book, pamphlet and periodical sources, as well as unpublished manuscripts in the National Library of Scotland and the Shetland Archives, this thesis breaks new ground in its investigation of the role of Scots verse translation in the second-generation Scottish Renaissance.

A Fuzzy Decision Strategy for Topic Identification and Dynamic Selection of Language Models

International audience

Integration of a graph-based model indexer in commercial modelling tools

Softeam has over 20 years of experience providing UML-based modelling solutions, such as its Modelio modelling tool, and its Constellation enterprise model management and collaboration environment. Due to the increasing number and size of the models used by Softeam’s clients, Softeam joined the MONDO FP7 EU research project, which worked on solutions for these scalability challenges and produced the Hawk model indexer among other results. This paper presents the technical details and several case studies on the integration of Hawk into Softeam’s toolset. The first case study measured the performance of Hawk’s Modelio support using varying amounts of memory for the Neo4j backend. In another case study, Hawk was integrated into Constellation to provide scalable global querying of model repositories. Finally, the combination of Hawk and the Epsilon Generation Language was compared against Modelio for document generation: for the largest model, Hawk was two orders of magnitude faster.

Kinetic modelling of in vitro data of PI3K, mTOR1, PTEN enzymes and on-target inhibitors Rapamycin, BEZ235, and LY294002

The phosphatidylinositide 3-kinases (PI3K) and mammalian target of rapamycin-1 (mTOR1) are two key targets for anti-cancer therapy. Predicting the response of the PI3K/AKT/mTOR1 signalling pathway to targeted therapy is made difficult because of network complexities. Systems biology models can help explore those complexities but the value of such models is dependent on accurate parameterisation. Motivated by a need to increase accuracy in kinetic parameter estimation, and therefore the predictive power of the model, we present a framework to integrate kinetic data from enzyme assays into a unified enzyme kinetic model. We present exemplar kinetic models of PI3K and mTOR1, calibrated on in vitro enzyme data and founded on Michaelis-Menten (MM) approximation. We describe the effects of an allosteric mTOR1 inhibitor (Rapamycin) and ATP-competitive inhibitors (BEZ2235 and LY294002) that show dual inhibition of mTOR1 and PI3K. We also model the kinetics of phosphatase and tensin homolog (PTEN), which modulates sensitivity of the PI3K/AKT/mTOR1 pathway to these drugs. Model validation with independent data sets allows investigation of enzyme function and drug dose dependencies in a wide range of experimental conditions. Modelling of the mTOR1 kinetics showed that Rapamycin has an IC50 independent of ATP concentration and that it is a selective inhibitor of mTOR1 substrates S6K1 and 4EBP1: it retains 40% of mTOR1 activity relative to 4EBP1 phosphorylation and inhibits completely S6K1 activity. For the dual ATP-competitive inhibitors of mTOR1 and PI3K, LY294002 and BEZ235, we derived the dependence of the IC50 on ATP concentration that allows prediction of the IC50 at different ATP concentrations in enzyme and cellular assays. Comparison of the drug effectiveness in enzyme and cellular assays showed that some features of these drugs arise from signalling modulation beyond the on-target action and MM approximation and require a systems-level consideration of the whole PI3K/PTEN/AKT/mTOR1 network in order to understand mechanisms of drug sensitivity and resistance in different cancer cell lines. We suggest that using these models in systems biology investigation of the PI3K/AKT/mTOR1 signalling in cancer cells can bridge the gap between direct drug target action and the therapeutic response to these drugs and their combinations.

Modelling the generation of customised poetry in Jess

The present paper presents an application that composes formal poetry in Spanish in a semiautomatic interactive fashion. JASPER is a forward reasoning rule-based system that obtains from the user an intended message, the desired metric, a choice of vocabulary, and a corpus of verses; and, by intelligent adaptation of selected examples from this corpus using the given words, carries out a prose-to-poetry translation of the given message. In the composition process, JASPER combines natural language generation and a set of construction heuristics obtained from formal literature on Spanish poetry.

Towards Unstructured Knowledge Integration in Natural Language Processing

In the last decades, Artificial Intelligence has witnessed multiple breakthroughs in deep learning. In particular, purely data-driven approaches have opened to a wide variety of successful applications due to the large availability of data. Nonetheless, the integration of prior knowledge is still required to compensate for specific issues like lack of generalization from limited data, fairness, robustness, and biases. In this thesis, we analyze the methodology of integrating knowledge into deep learning models in the field of Natural Language Processing (NLP). We start by remarking on the importance of knowledge integration. We highlight the possible shortcomings of these approaches and investigate the implications of integrating unstructured textual knowledge. We introduce Unstructured Knowledge Integration (UKI) as the process of integrating unstructured knowledge into machine learning models. We discuss UKI in the field of NLP, where knowledge is represented in a natural language format. We identify UKI as a complex process comprised of multiple sub-processes, different knowledge types, and knowledge integration properties to guarantee. We remark on the challenges of integrating unstructured textual knowledge and bridge connections with well-known research areas in NLP. We provide a unified vision of structured knowledge extraction (KE) and UKI by identifying KE as a sub-process of UKI. We investigate some challenging scenarios where structured knowledge is not a feasible prior assumption and formulate each task from the point of view of UKI. We adopt simple yet effective neural architectures and discuss the challenges of such an approach. Finally, we identify KE as a form of symbolic representation. From this perspective, we remark on the need of defining sophisticated UKI processes to verify the validity of knowledge integration. To this end, we foresee frameworks capable of combining symbolic and sub-symbolic representations for learning as a solution.

A Unified Model for XVA, including Interest Rates and Rating

We start in Chapter 2 to investigate linear matrix-valued SDEs and the Itô-stochastic Magnus expansion. The Itô-stochastic Magnus expansion provides an efficient numerical scheme to solve matrix-valued SDEs. We show convergence of the expansion up to a stopping time τ and provide an asymptotic estimate of the cumulative distribution function of τ. Moreover, we show how to apply it to solve SPDEs with one and two spatial dimensions by combining it with the method of lines with high accuracy. We will see that the Magnus expansion allows us to use GPU techniques leading to major performance improvements compared to a standard Euler-Maruyama scheme. In Chapter 3, we study a short-rate model in a Cox-Ingersoll-Ross (CIR) framework for negative interest rates. We define the short rate as the difference of two independent CIR processes and add a deterministic shift to guarantee a perfect fit to the market term structure. We show how to use the Gram-Charlier expansion to efficiently calibrate the model to the market swaption surface and price Bermudan swaptions with good accuracy. We are taking two different perspectives for rating transition modelling. In Section 4.4, we study inhomogeneous continuous-time Markov chains (ICTMC) as a candidate for a rating model with deterministic rating transitions. We extend this model by taking a Lie group perspective in Section 4.5, to allow for stochastic rating transitions. In both cases, we will compare the most popular choices for a change of measure technique and show how to efficiently calibrate both models to the available historical rating data and market default probabilities. At the very end, we apply the techniques shown in this thesis to minimize the collateral-inclusive Credit/ Debit Valuation Adjustments under the constraint of small collateral postings by using a collateral account dependent on rating trigger.

Vocabulary Transfer and Knowledge Distillation for Language Model Compression

Natural Language Processing (NLP) has seen tremendous improvements over the last few years. Transformer architectures achieved impressive results in almost any NLP task, such as Text Classification, Machine Translation, and Language Generation. As time went by, transformers continued to improve thanks to larger corpora and bigger networks, reaching hundreds of billions of parameters. Training and deploying such large models has become prohibitively expensive, such that only big high tech companies can afford to train those models. Therefore, a lot of research has been dedicated to reducing a model’s size. In this thesis, we investigate the effects of Vocabulary Transfer and Knowledge Distillation for compressing large Language Models. The goal is to combine these two methodologies to further compress models without significant loss of performance. In particular, we designed different combination strategies and conducted a series of experiments on different vertical domains (medical, legal, news) and downstream tasks (Text Classification and Named Entity Recognition). Four different methods involving Vocabulary Transfer (VIPI) with and without a Masked Language Modelling (MLM) step and with and without Knowledge Distillation are compared against a baseline that assigns random vectors to new elements of the vocabulary. Results indicate that VIPI effectively transfers information of the original vocabulary and that MLM is beneficial. It is also noted that both vocabulary transfer and knowledge distillation are orthogonal to one another and may be applied jointly. The application of knowledge distillation first before subsequently applying vocabulary transfer is recommended. Finally, model performance due to vocabulary transfer does not always show a consistent trend as the vocabulary size is reduced. Hence, the choice of vocabulary size should be empirically selected by evaluation on the downstream task similar to hyperparameter tuning.

NLG-Metricverse: An End-to-End Library for Evaluating Natural Language Generation

Driven by recent deep learning breakthroughs, natural language generation (NLG) models have been at the center of steady progress in the last few years. However, since our ability to generate human-indistinguishable artificial text lags behind our capacity to assess it, it is paramount to develop and apply even better automatic evaluation metrics. To facilitate researchers to judge the effectiveness of their models broadly, we suggest NLG-Metricverse—an end-to-end open-source library for NLG evaluation based on Python. This framework provides a living collection of NLG metrics in a unified and easy- to-use environment, supplying tools to efficiently apply, analyze, compare, and visualize them. This includes (i) the extensive support of heterogeneous automatic metrics with n-arity management, (ii) the meta-evaluation upon individual performance, metric-metric and metric-human correlations, (iii) graphical interpretations for helping humans better gain score intuitions, (iv) formal categorization and convenient documentation to accelerate metrics understanding. NLG-Metricverse aims to increase the comparability and replicability of NLG research, hopefully stimulating new contributions in the area.

[adherence To Statin Treatment And Associated Factors In Female Users From The Unified Health System (sus)].

To identify the adherence rate of a statin treatment and possible related factors in female users from the Unified Health System. Seventy-one women were evaluated (64.2 ± 11.0 years) regarding the socio-economic level, comorbidities, current medications, level of physical activity, self-report of muscular pain, adherence to the medical prescription, body composition and biochemical profile. The data were analyzed as frequencies, Chi-Squared test, and Mann Whitney test (p<0.05). 15.5% of women did not adhere to the medical prescription for the statin treatment, whose had less comorbidities (p=0.01), consumed less quantities of medications (p=0.00), and tended to be younger (p=0.06). Those patients also presented higher values of lipid profile (CT: p=0.01; LDL-c: p=0.02). Musculoskeletal complains were not associated to the adherence rate to the medication. The associated factors to adherence of dyslipidemic women to statin medical prescription were age, quantity of comorbidities and quantity of current medication.