Recent studies have shown that strong Natural Language Understanding (NLU) models are prone to relying on annotation biases of the datasets as a shortcut, which goes against the underlying mechanisms of the task of interest. To reduce such biases, several recent works introduce debiasing methods to regularize the training process of targeted NLU models. In this paper, we provide a new perspective with causal inference to find out the bias. On one hand, we show that there is an unobserved confounder for the natural language utterances and their respective classes, leading to spurious correlations from training data. To remove such confounder, the backdoor adjustment with causal intervention is utilized to find the true causal effect, which makes the training process fundamentally different from the traditional likelihood estimation. On the other hand, in inference process, we formulate the bias as the direct causal effect and remove it by pursuing the indirect causal effect with counterfactual reasoning. We conduct experiments on large-scale natural language inference and fact verification benchmarks, evaluating on bias sensitive datasets that are specifically designed to assess the robustness of models against known biases in the training data. Experimental results show that our proposed debiasing framework outperforms previous state-of-the-art debiasing methods while maintaining the original in-distribution performance.

In order to curb the illegal activities in power resources detection, improve the credibility and contribution rate of the industry, and promote the development of high-quality services, this paper proposes a secure and reliable trusted testing system of electric power materials based on blockchain. Firstly, a device and personal information query authorization mechanism is established to provide solutions for personnel and testing equipment authorization. It can help ensure the reliability of testing data on the premise of security. Secondly, we propose a method to deal with the difficulties of testing information management. Lastly, we introduce the case of electricity management helping the power authorities to supervise effectively and increasing the credibility of power material procurement evidence to prove the feasibility of this system.

This paper proposed a non-filter position sensorless control (PSC) for the permanent-magnet synchronous motor (PMSM), which is based on an a-β frame complex PI controller. The chattering of estimated Back-EMFs is unavoidable with a sliding mode PSC. A low pass filter (LPF) is always employed to smooth Back-EMFs; however phase lag of LPF becomes significant as motor speed increases. A complex PI controller is proposed to estimate back-EMFs without LPFs. The proposed method is implemented by replacing SMC and cascaded LPFs with a single complex PI controller, therefore complexities of proposed PSC have not been increased. Experiments were carried out on a PMSM drive to confirm the advantages of the proposed method over sliding mode PSC.

Document classification is an essential task in many real world applications. Existing approaches adopt both text semantics and document structure to obtain the document representation. However, these models usually require a large collection of annotated training instances, which are not always feasible, especially in low-resource settings. In this paper, we propose a multi-task learning framework to jointly train multiple related document classification tasks. We devise a hierarchical architecture to make use of the shared knowledge from all tasks to enhance the document representation of each task. We further propose an inter-attention approach to improve the task-specific modeling of documents with global information. Experimental results on 15 public datasets demonstrate the benefits of our proposed model.

Electronic Health Records (EHRs) store a large amount of clinical data associated with each patient.Information extraction on unstructured clinical notes in EHRs is important which could contribute to huge improvement in patient health management.Previous studies mainly focused on English corpus.However, at the same time there are very limited research work on Chinese EHRs.Due to the challenges brought by the characteristics of Chinese, it is difficult to apply existing techniques for English on Chinese corpus.In this paper, we propose a deep learning based framework for information extraction from clinical notes in Chinese EHRs.Our framework consists of three components: data preprocessing, feature generation and entity and relation extractor.For clinical entity recognition, we propose a novel Conditional Random Field (CRF) based model and introduce effective features by leveraging the characteristics of Chinese language.For relation extraction, we utilize Convolutional Neural Network (CNN)to obtain high quality entity-relation facts.To the best of our knowledge, this is the first framework to apply deep learning to information extraction from clinical notes in Chinese EHRs.We conduct extensive sets of experiments on real-world datasets from hospital.The experimental results show the effectiveness of our framework, indicating its practical application value.

A novel nonlinear model for surface permanent magnet synchronous motors (SPMSMs) is adopted to estimate the initial rotor position for hybrid electric vehicles (HEVs). Usually, the accuracy of initial rotor position estimation for SPMSMs relies on magnetic saturation. To verify the saturation effect, the transient finite element analysis (FEA) model is presented first. Hybrid injection of a static voltage vector (SVV) superimposed with a high-frequency rotating voltage is proposed. The magnetic polarity is roughly identified with the aid of the saturation evaluation function, based on which an estimation of the position is performed. During this procedure, a special demodulation is suggested to extract signals of iron core saturation and rotor position. A Simulink/MATLAB platform for SPMSMs at standstill is constituted, and the effectiveness of the proposed strategy is verified. The proposed method is also validated by experimental results of an SPMSM drive.

We present a hidden Markov modeling (HMM) based approach for recognition of articulated objects in synthetic aperture radar (SAR) images. We develop multiple models for a given SAR image of an object and integrate these models synergistically using their probabilistic estimates for recognition and estimates of invariance of features as a result of articulation. The models are based on sequentialization of scattering centers extracted from SAR images. Experimental results are presented using 1440 training images and 2520 testing images for 4 classes.