基于卷积神经网络的词义消歧

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.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.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.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.实验结果表明，该方法在词义消歧中是可行的.

Word Sense Disambiguation Based on Convolution Neural Network

In order to improve the performance of word sense disambiguation (WSD), a disambiguation method based on convolution neural network (CNN) is proposed. Ambiguous word is viewed as center and four adjacent word units around its left and right sides are extended. Word, part-of-speech and semantic categories are extracted as disambiguation features. Based on disambiguation features, CNN is used to determine semantic categories of ambiguous words. Training corpus of SemEval-2007:Task#5 and 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. Experiments show that this method is feasible in WSD.