文献类型：期刊文献

英文题名：Research on the statistical machine translation based on neural network learning

作者：Wang, Xueling[1]

第一作者：王雪玲

通讯作者：Wang, Xueling

机构：[1] School of Foreign Languages, Xinxiang University, Xinxiang, 453003, China

第一机构：新乡学院外国语学院

年份：2017

卷号：32

期号：3

起止页码：757-766

外文期刊名：Revista de la Facultad de Ingenieria

收录：EI(收录号：20173003966406)

语种：英文

外文关键词：Character recognition - Computational linguistics - Computer aided language translation - Feature extraction - Knowledge representation - Learning algorithms - Linguistics - Natural language processing systems - Parameter estimation - Speech recognition - Translation (languages)

摘要：In order to solve the problem of word alignment in foreign language translation, the principle of neural network learning model, parameters estimation and feature selection algorithm are investigated. Neural network learning model is an effective method to set up statistical language model, which has the strong knowledge representation ability. The training speed of neural network learning method is very slow, thus it consumes a lot of resources. Because the feature space is commonly very large, how to choose typical and less redundant characteristics for model training is very important. Then an improved feature selection algorithm is put forward. Word alignment system based on neural network learning model is established, which includes text pre-processing, model training, name entity recognition and part of speech tagging. At last, experiments are done to test its performance. The results show that the proposed system has high accuracy. ? 2017 Universidad Central de Venezuela.