Recurrent neural networks (RNN) used for Chinese named entity recognition (NER) that sequentially track character and word information have achieved great success. However, the characteristic of chain structure and the lack of global semantics determine that RNN-based models are vulnerable to word ambiguities. In this work, we try to alleviate this problem by introducing a lexicon-based graph neural network with global semantics, in which lexicon knowledge is used to connect characters to capture the local composition, while a global relay node can capture global sentence semantics and longrange dependency. Based on the multiple graph-based interactions among characters, potential words, and the whole-sentence semantics, word ambiguities can be effectively tackled. Experiments on four NER datasets show that the proposed model achieves significant improvements against other baseline models.

Advanced modulation formats combined with digital signal processing and direct detection is a promising way to realize high capacity, low cost and power efficient short reach optical transmission system. In this paper, we present a detailed investigation on the performance of three advanced modulation formats for 100 Gb/s short reach transmission system. They are PAM-4, CAP-16 and DMT. The detailed digital signal processing required for each modulation format is presented. Comprehensive simulations are carried out to evaluate the performance of each modulation format in terms of received optical power, transmitter bandwidth, relative intensity noise and thermal noise. The performance of each modulation format is also experimentally studied. To the best of our knowledge, we report the first demonstration of a 112 Gb/s transmission over 10km of SSMF employing single band CAP-16 with EML. Finally, a comparison of computational complexity of DSP for the three formats is presented.

Character-level Chinese named entity recognition (NER) that applies long short-term memory (LSTM) to incorporate lexicons has achieved great success. However, this method fails to fully exploit GPU parallelism and candidate lexicons can conflict. In this work, we propose a faster alternative to Chinese NER: a convolutional neural network (CNN)-based method that incorporates lexicons using a rethinking mechanism. The proposed method can model all the characters and potential words that match the sentence in parallel. In addition, the rethinking mechanism can address the word conflict by feeding back the high-level features to refine the networks. Experimental results on four datasets show that the proposed method can achieve better performance than both word-level and character-level baseline methods. In addition, the proposed method performs up to 3.21 times faster than state-of-the-art methods, while realizing better performance.

In this work, we study the problem of partof-speech tagging for Tweets. In contrast to newswire articles, Tweets are usually informal and contain numerous out-ofvocabulary words. Moreover, there is a lack of large scale labeled datasets for this domain. To tackle these challenges, we propose a novel neural network to make use of out-of-domain labeled data, unlabeled in-domain data, and labeled indomain data. Inspired by adversarial neural networks, the proposed method tries to learn common features through adversarial discriminator. In addition, we hypothesize that domain-specific features of target domain should be preserved in some degree. Hence, the proposed method adopts a sequence-to-sequence autoencoder to perform this task. Experimental results on three different datasets show that our method achieves better performance than state-of-the-art methods.

Long-haul optical communications based on nonlinear Fourier Transform(NFT) has gained attention recently as a new communication strategy that inherently embrace the nonlinear nature of the optical fiber. For communications using discrete eigenvalues 𝝀 ∈ ℂ + , information are encoded and decoded in the spectral amplitudes 𝒒(𝝀) = 𝒃(𝝀) ( 𝒅𝒂(𝝀) 𝒅𝝀 ) ⁄ at the root 𝝀 𝒓𝒕 where 𝒂(𝝀 𝒓𝒕 ) = 𝟎 . In this paper, we propose two alternative decoding methods using 𝒂(𝝀) and 𝒃(𝝀) instead of 𝒒(𝝀) as decision metrics. For discrete eigenvalue modulation systems, we show that symbol decisions using 𝒂(𝝀) at a prescribed set of 𝝀 values perform similarly to conventional methods using 𝒒(𝝀) but avoids root searching and thus significantly reduce computational complexity. For systems with phase and amplitude modulation on a given discrete eigenvalue, we propose to use 𝒃(𝝀) after for symbol detection and show that the noise in 𝒅𝒂(𝝀) 𝒅𝝀 and 𝝀 𝒓𝒕 after transmission are all correlated with that in 𝒃(𝝀 𝒓𝒕 ) . A linear minimum mean square error(LMMSE) estimator of the noise in 𝒃(𝝀 𝒓𝒕 ) is derived based on such noise correlation and transmission performance is considerably improved for QPSK and 16-QAM systems on discrete eigenvalues.