| ID | 原文 | 译文 |
| 58098 | 最后对所有胶囊的输出归一化，得到两段文本的蕴含关系. | Finally，the output of all capsules is normalized to obtain the relationship between the twotexts. |
| 58099 | 该方法在 SNLI 测试集上的准确率为 89. | The accuracy on SNLI test dataset is 89. |
| 58100 | 2% ，在 MultiNLI 匹配测试集和不匹配测试集上的准确率分别为 77. | 2% ，the accuracy on MultiNLI matched and mismatchedtest dataset is 77. |
| 58101 | 4% 和 76.4% . | 4% and 76. 4% . |
| 58102 | 对交互模块中注意力关系矩阵的可视化分析结果验证了胶囊在英文文本蕴含识别任务中的有效性. | The visual analysis of the attentional relationship matrix of interactionmodule also verifies the effectiveness of capsules in the English textual entailment recognition task. |
| 58103 | 为了提高词义消歧性能，提出了一种基于卷积神经网络的消歧方法. | In order to improve the performance of word sense disambiguation ( WSD) ，a disambiguationmethod based on convolution neural network ( CNN) is proposed. |
| 58104 | 以歧义词为中心，向左右两侧连续扩展4 个邻接词汇单元，选取其中的词形、词性和语义类作为消歧特征. | Ambiguous word is viewed as centerand four adjacent word units around its left and right sides are extended. Word，part-of-speech and semantic categories are extracted as disambiguation features. |
| 58105 | 以消歧特征为基础，使用卷积神经网络来确定歧义词的语义类别. | Based on disambiguation features，CNN isused to determine semantic categories of ambiguous words. |
| 58106 | 利用 SemEval-2007:Task#5 的训练语料和哈尔滨工业大学语义标注语料来优化卷积神经网络.使用 SemEval-2007:Task#5 的测试语料来测试词义消歧分类器的性能，所提方法的消歧平均准确率有提高. | Training corpus of SemEval-2007: Task#5and semantic annotation corpus from Harbin Institute of Technology are used to optimize CNN classifier. Testing corpus of SemEval-2007: Task#5 is used to test the performance of WSD classifier. Average disambiguation accuracy of the proposed method is improved. |
| 58107 | 实验结果表明，该方法在词义消歧中是可行的. | Experiments show that this method is feasiblein WSD. |