The research on named entity recognition for label-few domain is becoming increasingly important.In this paper,a novel algorithm,positive unlabeled named entity recognition(PUNER)with multi-granularity language inform...The research on named entity recognition for label-few domain is becoming increasingly important.In this paper,a novel algorithm,positive unlabeled named entity recognition(PUNER)with multi-granularity language information,is proposed,which combines positive unlabeled(PU)learning and deep learning to obtain the multi-granularity language information from a few labeled in-stances and many unlabeled instances to recognize named entities.First,PUNER selects reliable negative instances from unlabeled datasets,uses positive instances and a corresponding number of negative instances to train the PU learning classifier,and iterates continuously to label all unlabeled instances.Second,a neural network-based architecture to implement the PU learning classifier is used,and comprehensive text semantics through multi-granular language information are obtained,which helps the classifier correctly recognize named entities.Performance tests of the PUNER are carried out on three multilingual NER datasets,which are CoNLL2003,CoNLL 2002 and SIGHAN Bakeoff 2006.Experimental results demonstrate the effectiveness of the proposed PUNER.展开更多
Graph embedding aims to map the high-dimensional nodes to a low-dimensional space and learns the graph relationship from its latent representations.Most existing graph embedding methods focus on the topological struct...Graph embedding aims to map the high-dimensional nodes to a low-dimensional space and learns the graph relationship from its latent representations.Most existing graph embedding methods focus on the topological structure of graph data,but ignore the semantic information of graph data,which results in the unsatisfied performance in practical applications.To overcome the problem,this paper proposes a novel deep convolutional adversarial graph autoencoder(GAE)model.To embed the semantic information between nodes in the graph data,the random walk strategy is first used to construct the positive pointwise mutual information(PPMI)matrix,then,graph convolutional net-work(GCN)is employed to encode the PPMI matrix and node content into the latent representation.Finally,the learned latent representation is used to reconstruct the topological structure of the graph data by decoder.Furthermore,the deep convolutional adversarial training algorithm is introduced to make the learned latent representation conform to the prior distribution better.The state-of-the-art experimental results on the graph data validate the effectiveness of the proposed model in the link prediction,node clustering and graph visualization tasks for three standard datasets,Cora,Citeseer and Pubmed.展开更多
The performance of detector limits the overall performance of laser ranging system. And the design of multi-hit detector is one of the feasible ways to promote the performance of detector. Currently, the segmentation ...The performance of detector limits the overall performance of laser ranging system. And the design of multi-hit detector is one of the feasible ways to promote the performance of detector. Currently, the segmentation method or the recursive method is commonly used to analyze the multi-hit detector model. To the best of our knowledge, this paper is the first to propose a combinatorial method to solve the multi-hit detector model from the perspective of discrete time. Then, universal formulas of total signal detection probability and the average count are deduced based on the Poisson distribution signal. Furthermore, analysis is made to figure out how the average count changes with different parameters, such as the dead time, gating time, rate intensity. As a result, for GM-APD, the multi-hit detector model is verified advantageously compared to the single-hit detector model in improving the average count theoretically. Meanwhile, a discrete step feature is presented when average count changes with dead time or the gating time, which is of great significance in gating time optimization.展开更多
This Letter introduces a trigger-controlled Geiger-mode avalanche photodiode （GM-APD）. A hierarchical look- back-upon tree recurrence method is given to predict the performance of trigger-controlled GM-APDs under di...This Letter introduces a trigger-controlled Geiger-mode avalanche photodiode （GM-APD）. A hierarchical look- back-upon tree recurrence method is given to predict the performance of trigger-controlled GM-APDs under different trigger-count upper limits. In addition, the normalized detection probability is defined to evaluate the detection performance of trigger-controlled GM-APDs in typical weak optical signal detection （impulse noise and continuous noise situations）. Theoretical analyses show that the trigger-controlled GM-APD improves the detection performance of GM-APDs in weak optical signal detection via the optimization of the trigger-count upper limit, compared with single-trigger and multi-trigger GM-APDs.展开更多
基金the National Natural Science Foundation of China(No.61876144)the Strategy Priority Research Program of Chinese Acade-my of Sciences(No.XDC02070600).
文摘The research on named entity recognition for label-few domain is becoming increasingly important.In this paper,a novel algorithm,positive unlabeled named entity recognition(PUNER)with multi-granularity language information,is proposed,which combines positive unlabeled(PU)learning and deep learning to obtain the multi-granularity language information from a few labeled in-stances and many unlabeled instances to recognize named entities.First,PUNER selects reliable negative instances from unlabeled datasets,uses positive instances and a corresponding number of negative instances to train the PU learning classifier,and iterates continuously to label all unlabeled instances.Second,a neural network-based architecture to implement the PU learning classifier is used,and comprehensive text semantics through multi-granular language information are obtained,which helps the classifier correctly recognize named entities.Performance tests of the PUNER are carried out on three multilingual NER datasets,which are CoNLL2003,CoNLL 2002 and SIGHAN Bakeoff 2006.Experimental results demonstrate the effectiveness of the proposed PUNER.
基金Supported by the Strategy Priority Research Program of Chinese Academy of Sciences(No.XDC02070600).
文摘Graph embedding aims to map the high-dimensional nodes to a low-dimensional space and learns the graph relationship from its latent representations.Most existing graph embedding methods focus on the topological structure of graph data,but ignore the semantic information of graph data,which results in the unsatisfied performance in practical applications.To overcome the problem,this paper proposes a novel deep convolutional adversarial graph autoencoder(GAE)model.To embed the semantic information between nodes in the graph data,the random walk strategy is first used to construct the positive pointwise mutual information(PPMI)matrix,then,graph convolutional net-work(GCN)is employed to encode the PPMI matrix and node content into the latent representation.Finally,the learned latent representation is used to reconstruct the topological structure of the graph data by decoder.Furthermore,the deep convolutional adversarial training algorithm is introduced to make the learned latent representation conform to the prior distribution better.The state-of-the-art experimental results on the graph data validate the effectiveness of the proposed model in the link prediction,node clustering and graph visualization tasks for three standard datasets,Cora,Citeseer and Pubmed.
文摘The performance of detector limits the overall performance of laser ranging system. And the design of multi-hit detector is one of the feasible ways to promote the performance of detector. Currently, the segmentation method or the recursive method is commonly used to analyze the multi-hit detector model. To the best of our knowledge, this paper is the first to propose a combinatorial method to solve the multi-hit detector model from the perspective of discrete time. Then, universal formulas of total signal detection probability and the average count are deduced based on the Poisson distribution signal. Furthermore, analysis is made to figure out how the average count changes with different parameters, such as the dead time, gating time, rate intensity. As a result, for GM-APD, the multi-hit detector model is verified advantageously compared to the single-hit detector model in improving the average count theoretically. Meanwhile, a discrete step feature is presented when average count changes with dead time or the gating time, which is of great significance in gating time optimization.
文摘This Letter introduces a trigger-controlled Geiger-mode avalanche photodiode （GM-APD）. A hierarchical look- back-upon tree recurrence method is given to predict the performance of trigger-controlled GM-APDs under different trigger-count upper limits. In addition, the normalized detection probability is defined to evaluate the detection performance of trigger-controlled GM-APDs in typical weak optical signal detection （impulse noise and continuous noise situations）. Theoretical analyses show that the trigger-controlled GM-APD improves the detection performance of GM-APDs in weak optical signal detection via the optimization of the trigger-count upper limit, compared with single-trigger and multi-trigger GM-APDs.
