ID	原文	译文
47086	最后，通过 mean-pooling 层处理句子向量后且使用了 softmax 层进行文本分类。	Finally, the sentence vectors were processed by mean-pooling layer and text categorization was clas-sified by softmax layer.
47087	实验验证了 Bi-LSTM 和 double word-embedding 神经网络相结合的模型训练效果与提取情况。	The training effects and extraction performance of the combination model of Bi-LSTM anddouble word-embedding neural network were verified.
47088	实验结果表明，该模型不但能较好地处理高质量的文本特征向量提取和表达序列，而且比 LSTM、LSTM+context window 和 Bi-LSTM这 3 种神经网络有较明显的表达效果。	The experimental results show that this model not only performswell in dealing with the high-quality text feature vector and the expression sequence, but also significantly outperformsother three kinds of neural networks, which includes LSTM, LSTM+context window and Bi-LSTM.
47089	Tanner 图中的环分布影响着低密度校验码（LDPC, low-density parity-check code）译码算法的误码率性能，	Loop distribution of Tanner graph affects the BER performance of low-density parity-check codes(LDPC) de-coding.
47090	为快速计算出 Tanner 图中短环的数目，提出一种逐边递推基于矩阵运算的算法。	To count short cycles in the Tanner graph efficiently, a side by side recursion algorithm based on matrix computa-tion was proposed.
47091	首先定义 5 种基本图结构，算法在实施过程中可实现结构间的递推。	Firstly, 5 basic graph structures were defined to realize recursive calculate in the implementationprocess.
47092	与之前的研究工作相比，该算法对于同一环长提供多种方法进行计算，	Compared with previous works, the algorithm provided many methods for counting the same length of cycles.
47093	得到相同的计算结果，进一步证实算法的正确性。	The same result confirmed the correctness of the algorithm.
47094	新算法不仅能计算出总的环数，还能给出每一条边参与的环数。	The new algorithm could not only calculate the total numberof cycles, but also gave the number each edge participating in fixed-length cycles.
47095	该算法将时间复杂度从正比于码长 N 的 3 次方降为正比于码长的平方与变量节点平均度数 D 的乘积（D<	Its complexity was proportional to theproduct of D and square of N, where D was the average degree of variable nodes, and N denoted the code length. ForLDPC codes, D was far less than N.