Search

The BERT pre-trained language model removes the tones of the syllables when segmenting Vietnamese words,which leads to the loss of some semantic information during the training process of grammatical error detection model.To address this problem,an approach combining part of speech and tonal features is proposed to complete the semantic information of the input syllables.Grammatical error detection task confronts the problem of insufficient training data due to the scarcity of labeled Vietnamese data.To address this problem,a data augmentation algorithm is designed to generate a large number of error texts from the correct corpus.Experimental results on Vietnamese Wikipedia and news corpus show that the proposed method achieves the highest F0.5 and F1 score on the test set,which proves it improves the detection performance.Both the proposed method and the baseline model method have a gradual improvement with the increasing scales of the generated data,which proves that the proposed data augmentation algorithm is effective.

Dependency parsing aims to identify syntactic dependencies between words in a sentence.Dependency parsing can provide syntactic features and improve model performance for tasks such as information extraction,automatic question answering and machine translation.The training data size has an significant impact on the performance of the dependency parsing model.The lack of training data will cause serious unknown word problems and model over-fitting problems.This paper proposes various data augment strategies for the problem of low-resource dependency parsing.The proposed method effectively expands the training data by synonym substitution and alleviates the unknown words problem.The data augment strategies of multiple Mixups effectively alleviate the model overfitting problem and improve the generalization ability of the model.Experimental results on the universal dependencies treebanks(UD treebanks) dataset show that the proposed methods effectively improve the performance of Thai,Vietnamese and English dependency parsing under small-scale training corpus conditions.

Aspect extraction of the microblog involved in the case is a task in a specific domain.The expression of aspect words is diverse and the meaning is different from that of the general domain.Only relying on the word embedding in the general domain,these aspect words cannot be well represented.This paper proposes a method for extracting aspect words from microblogs by using both domain word embedding and generic word embedding.Firstly,all the microblogs involved in the case is pre-trained to obtain the embedding layer with the characteristics of the involved domain.Secondly,the microblog comments are input into two embedding layers to obtain the characterization results of the aspect words in different domains,and perform the splicing operation.Then,the features related to the case are extracted through the convolution layer.Finally,the classifier is used to label the sequence to extract aspect words involved in the case.The experimental results show that the F1 value of the proposed method reaches 72.36% and 71.02% respectively on the data sets of #Chongqing bus falling into the river# and #Mercedes Benz female driver rights protection#,which is better than the existing benchmark models,and verifies the influence of word embedding in different domains on the aspect extraction of the microblogs.

Aiming at the inaccurate phoneme segmentation problem caused by the lack of consideration of Lao language tone changes and audio diversity in existing methods, this paper proposes an unsupervised phoneme segmentation method for Lao language with multi-feature interaction fusion. Firstly, self-supervised features, spectral features and pitch features are independently coded to avoid the insufficiency of a single feature. Secondly, multiple independent features are gradually fused based on the attention mechanism, so that the model can more comprehensively capture the information of Lao language tone changes and phoneme boundaries. Finally, a learnable framework is adopted to optimize the phoneme segmentation model. The experimental results show that the proposed method improves the R-value by 27.88% on the Lao phoneme segmentation task compared with the baseline methods. [ABSTRACT FROM AUTHOR]

The Songjianghe deposit is a newly discovered altered gold deposit in the southeastern Jiapigou-Haigou Gold Metallogenic Belt (JHGMB) in southeastern Jilin Province of NE China. The host rocks were considered to be the Mesoproterozoic Seluohe Group, and the metallogenic epoch lacked accurate isotopic constraints. To determine the age and metallogenic setting of the deposit, we describe the geologic characteristics of the deposit and present the results of petrographic and geochronologic analyses of the host rocks and ores. The ore bodies are hosted within a suite of amphibolite facies metamorphic rocks superimposed by greenschist facies indicative of retrograde metamorphism. Zircon U--Pb dating results indicate that the host rocks belong to the Jiapigou Group that formed at the end of the Neoarchean (2543--2527 Ma). Subsequently, the rocks successively underwent metamorphism during the late Neoarchean (2521--2506 Ma), retrograde metamorphism caused by the closure of the Paleo-Asian Ocean during the late Permian to Early Triassic (262--250 Ma), and extension after the closure of the Paleo-Asian Ocean during the Late Triassic (231--210 Ma). Sericite [.sup.40]Ar/[.sup.39]Ar dating results suggest that the Songjianghe deposit formed during the Late Jurassic between 157 Ma and 156 Ma. By combining these new insights with those of previous studies, we propose that the Songjianghe deposit is a mesothermal gold deposit and that mineralization occurred during the extensional period in the intermittent stage that followed the first subduction of the Paleo-Pacific Plate. All the gold deposits in the JHGMB formed from the late Permian to Early Cretaceous by multi-stage mineralization events that corresponded temporally with the tectonic evolution of the Paleo-Asian Ocean and the episodic subduction of the Paleo-Pacific Plate. Key words: geochronology, sericite [.sup.40]Ar/[.sup.39]Ar, mesothermal gold deposit, Songjianghe, Northeast China. Le gisement de Songjianghe est un gisement aurifere altere decouvert recemment dans le sud-est de la ceinture metallogenique aurifere de Jiapigou-Haigou (CMAJH), dans le sud-est de la province de Jilin du nord-est de la Chine. Les roches hôtes etaient considerees comme constituant le Groupe mesoproterozoïque de Seluohe, et l'epoque metallogenique n'etait pas delimitee par des donnees isotopiques exactes. Afin de determiner l'age et le contexte metallogenique du gisement, nous en decrivons les caracteristiques geologiques et presentons les resultats d'analyses petrographiques et geochronologiques des roches hôtes et des minerais. Les corps mineralises se trouvent dans une serie de roches metamorphiques au facies des amphibolites sur lequel est surimpose un metamorphisme retrograde au facies des schistes verts. Les resultats de la datation U-Pb sur zircon indiquent que les roches hôtes appartiennent au Groupe de Jiapigou qui s'est forme a la fin du Neoarcheen (2543--2527 Ma). Par la suite, les roches ont successivement subi un metamorphisme durantle Neoarcheen tardif (2521--2506 Ma), un metamorphisme retrograde cause parlafermeture de l'ocean Paleoasiatique du Permien tardif au Trias precoce (262--250 Ma) et une extension apres la fermeture de l'ocean Paleoasiatique durant le Trias tardif (231-210 Ma). Les resultats de la datation [.sup.40]Ar/[.sup.39]Ar sur sericite donnent a penser que le gisement de Songjianghe s'est forme au Jurassique tardif, entre 157 Ma et 156 Ma. En combinant ces nouveaux renseignements aux resultats d'etudes anterieures, nous proposons que le gisement de Songjianghe est un gisement aurifere mesothermal et que la mineralisation s'est produite durant la periode d'extension dans l'intervalle qui a suivi la premiere subduction de la plaque paleo-Pacifique. Tous les gisements auriferes de la CMAJH se sont formes du Permien tardif au Cretace precoce par des episodes de mineralisation en plusieurs etapes qui correspondaient dans le temps avec l'evolution tectonique de l'ocean Paleoasiatique et la subduction episodique de la plaque paleo-Pacifique. [Traduit par la Redaction] Mots-cles: geochronologie, [.sup.40]Ar/[.sup.39]Ar sur sericite, gisement aurifere mesothermal, Songjianghe, nord-est de la Chine., 1. Introduction The Jiapigou-Haigou Gold Metallogenic Belt (JHGMB) is located in the NW-striking collage zone between the eastern section ofthe northern margin of the North China Craton (NCC) and the [...]

Text normalization (TN) is an indispensable work in the front-end analysis task of speech synthesis text. Lao text normalization is to convert non-standard words (NSW) in Lao text into spoken-form words (SFW). At present, the task of text normalization has not yet been carried out in Lao, which mainly faces the problems of difficult acquisition of training data, diversified language expression and text regularization with ambiguity. A text normalization task in Lao is carried out. This ask is completed as a sequence tagging task, and neural networks are used to predict NSW with ambiguity in combination with context. The corpus of the Lao text normalization task is constructed, the model results is predicted through the neural network, the self-attention mechanism is increased to deepen the relationship between the sequence characters, and different strategies are explored to introduce the pre-trained language model. An accuracy of 67.59% is achieved on the test set. [ABSTRACT FROM AUTHOR]

Ischemic stroke (IS) has become a cerebrovascular disease of widespread concern. Overexpression of circUCK2 alleviates neuronal damage in IS. However, the specific regulatory mechanisms of circUCK2 are not fully understood. In this study, we found that circUCK2 and HECT domain E3 ubiquitin ligase 1 (HECTD1) were downregulated in IS models in vitro and in vivo. Overexpression of circUCK2 or HECTD1 inhibited endothelial‐mesenchymal transition (EndoMT) and protected the blood–brain barrier (BBB) in transient middle cerebral artery occlusion mice from damage. It was further discovered that circUCK2 regulated HECTD1 expressions by interacting with fused in sarcoma (FUS). Moreover, FUS overexpression partially restored the effect of circUCK2 on EndoMT, and overexpression of HECTD1 weakened the effect of FUS on EndoMT. Collectively, circUCK2 upregulates the expression of HECTD1 by combining with FUS and inhibits EndoMT to alleviate BBB damage in IS both in vivo and in vitro. [ABSTRACT FROM AUTHOR]

The Changfagou Cu deposit is a newly discovered porphyry deposit located in the southern Jilin Province of Northeastern China, on the northeastern margin of the North China Craton. To better understand the formation of the Cu deposit, we report the zircon U–Pb and molybdenite Re–Os dating, and Sr-, Nd-, and Hf- isotopic data of the granite porphyry. LA-ICP-MS dating of zircon grains from two mineral zones in the granite porphyry yield ages of 115.7 ± 0.8 and 115.3 ± 0.6 Ma, which is interpreted as the emplacement age of the granite porphyry. The molybdenite Re–Os model ages of 112.5 to 113.8 Ma, an isochron age of 113.3 ± 1.3 Ma, and a weighted mean model age of 113.0 ± 0.7 Ma, which represents the age of the Cu mineralization quite well. The Changfagou granite porphyry samples lack amphibole and muscovite, and are compositionally characterized by high SiO2, high Na2O+K2O, and low P2O5, enriched in some Rb, Th, U, and Pb, and depleted in Nb, Ta, Ti, P, and Eu. Mineralogical and geochemical features suggest that the Changfagou granite porphyry samples are slightly peraluminous and are of highly fractionated I-type granitoids. The granitic rocks also have relatively high (87Sr/86Sr)i(0.71199 to 0. 71422), and both low εNd(t) (−14.56 to −13.19) and εHf(t) values (−14.916 to −8.644), which suggest that Changfagou granite porphyry are derived from mixed sources of crustal and mantle, and diagenesis and mineralization were possibly related to the switch in subduction direction of the Palaeo-Pacific Plate in the late phase of Early Cretaceous. [ABSTRACT FROM PUBLISHER]

An algorithm for Chinese domain term extraction based on language feature is proposed. Domain terms in Chinese have three features: domain cohesiveness, domain relevancy and domain consensus. The algorithm to extract domain term integrates three statistical models which compute domain cohesiveness, domain relevancy and domain consensus respectively. Experimental results show that the algorithm has higher precision and recall than the method based on mutual information and log likelihood. An automatic evaluation method based on perplexity attenuation ratio is proposed, and the above algorithms are measured by the automatic evaluation method. [ABSTRACT FROM AUTHOR]

Terahertz spectrum is sensitive to the change of the nonlocal molecular vibration mode. Accordingly, the spectral waveform is susceptible to variety of physical and chemical factors, which will lead to peak changes, frequency shifts, and even deformation of the overall waveform. Component analysis and material identification from the correspondence between the fixed peak features and materials will prone to cause errors or mistakes. Therefore, to solve this problem, we proposed a method based on Kernel Optimization Relevance Vector Machine (KO-RVM), which extracts global graphic features to distinct from the local features extraction method. And we use Support Vector Regression (SVR) algorithm as comparison. The result shows that, when basis functions’ parameters of RVM are optimized with expectation-maximization algorithm, it will be suitable for feature extraction of terahertz transmission spectrum. The spectrum can be sparsely represented, and the amount of extracted graphic features is substantially reduced. Reconstruction models based on these features are capable of retaining the overall spectral characteristics, and fitting results for each band are more consistent, while the extracted spectrum features can be used as basis of similarity measurement and the common characteristics investigation between different materials.

Feature extraction and classification are the key issues of terahertz spectroscopy identification. Because many materials have no apparent absorption peaks in the terahertz band, it is difficult to extract theirs terahertz spectroscopy feature and identify. To this end, a novel of identify terahertz spectroscopy approach with Deep Belief Network (DBN) was studied in this paper, which combines the advantages of DBN and K-Nearest Neighbors (KNN) classifier. Firstly, cubic spline interpolation and S-G filter were used to normalize the eight kinds of substances (ATP, Acetylcholine Bromide, Bifenthrin, Buprofezin, Carbazole, Bleomycin, Buckminster and Cylotriphosphazene) terahertz transmission spectra in the range of 0.9-6 THz. Secondly, the DBN model was built by two restricted Boltzmann machine (RBM) and then trained layer by layer using unsupervised approach. Instead of using handmade features, the DBN was employed to learn suitable features automatically with raw input data. Finally, a KNN classifier was applied to identify the terahertz spectrum. Experimental results show that using the feature learned by DBN can identify the terahertz spectrum of different substances with the recognition rate of over 90%, which demonstrates that the proposed method can automatically extract the effective features of terahertz spectrum. Furthermore, this KNN classifier was compared with others (BP neural network, SOM neural network and RBF neural network). Comparisons showed that the recognition rate of KNN classifier is better than the other three classifiers. Using the approach that automatic extract terahertz spectrum features by DBN can greatly reduce the workload of feature extraction. This proposed method shows a promising future in the application of identifying the mass terahertz spectroscopy.

In the present paper, support vector machine (SVM) based on convex combination kernel function will be used for classification of THz pulse transmission spectra. Wavelet transform is used in data pre-processing. Peaks and valleys are regarded as location features of THz pulse transmission spectra, which are injected into maximum interval features of term frequency-inverse document frequency (TF-IDF). We can conclude weight of each sampling point from the information theory. The weight represents the possibility that sampling point becomes feature. According to the situation that different terahertz-transmission spectra are lack of obvious features, we composed a SVM classification model based on convex combination kernel function. Evaluation function should be used as an evaluation method for obtaining the parameters of optimal convex combination to achieve a better accuracy. When the optimal parameter of kenal founction was determined, we should compose the model for process of classification and prediction. Compared with the single kernel function, the method can be combined with transmission spectroscopic features with classification model iteratively. Thanks to the dimensional mapping process, outstanding margin of features can be gained for the samples of different terahertz transmission spectrum. We carried out experiments using different samples The results demonstrated that the new approach is on par or superior in terms of accuracy and much better in feature fusion than SVM with single kernel function.