In this paper, we study automatic keyphrase generation.Although conventional approaches to this task show promising results, they neglect correlation among keyphrases, resulting in duplication and coverage issues. To solve these problems, we propose a new sequence-to-sequence architecture for keyphrase generation named CorrRNN, which captures correlation among multiple keyphrases in two ways. First, we employ a coverage vector to indicate whether the word in the source document has been summarized by previous phrases to improve the coverage for keyphrases. Second, preceding phrases are taken into account to eliminate duplicate phrases and improve result coherence. Experiment results show that our model significantly outperforms the state-of-the-art method on benchmark datasets in terms of both accuracy and diversity.

No abstract

No abstract

In order to reduce the computational complexity, and improve the pitch/roll estimation accuracy of the low-cost attitude heading reference system (AHRS) under conditions of magnetic-distortion, a novel linear Kalman filter, suitable for nonlinear attitude estimation, is proposed in this paper. The new algorithm is the combination of two-step geometrically-intuitive correction (TGIC) and the Kalman filter. In the proposed algorithm, the sequential two-step geometrically-intuitive correction scheme is used to make the current estimation of pitch/roll immune to magnetic distortion. Meanwhile, the TGIC produces a computed quaternion input for the Kalman filter, which avoids the linearization error of measurement equations and reduces the computational complexity. Several experiments have been carried out to validate the performance of the filter design. The results demonstrate that the mean time consumption and the root mean square error (RMSE) of pitch/roll estimation under magnetic disturbances are reduced by 45.9% and 33.8%, respectively, when compared with a standard filter. In addition, the proposed filter is applicable for attitude estimation under various dynamic conditions.

The design principle of the magnetic flux concentrators (MFCs) is studied by finite element method (FEM) and experiment. Firstly, the establishment process of the FEM model in Comsol Multiphysics is elaborated. Based on the simulation model, the magnetic amplification distributions of different shaped MFCs is compared, and the shape design principle of the MFC is obtained. Then, the influence of size parameters on the magnetic amplification of T, triangle, and trapezoid shaped MFCs is analyzed, and the size design principle of the MFCs is obtained. Additionally, the linear working range of the three MFCs is compared. To verify the simulation results, the MFCs are prepared using 1j79 permalloy. The experimental platform is established to acquire the magnetic gain under different magnetic fields. Based on the results, the linear working range of the MFCs is analyzed. And the influence of size parameters on the magnetic gain is tested. The experimental results are compared with the simulation ones to verify the validity of the model.