利用BP神经网络的中文词义消歧模型

词义消歧在自然语言处理中一直是一个难点问题，同时，也是很多领域都需要解决的一个重要环节．本文介绍了一种基于BP神经网络和统计方法相结合的有导词义消歧模型，详细讲解了BP神经网络原理；对使用这种混合人工智能的消歧模型的可能性和优越性进行了讨论；通过试验发现实际和预测结果的误差并不随着试验迭代次数而递减，而是实际误差随着次数的增加在零的附近呈现波动状态，即使用很少的迭代次数也可以得到比较好的结果．由此得出，BP神经网络预测模型在词义消歧的中具有良好的应用前景．

Using the BP neural networks to Chinese word sense disambiguation

Word sense disambiguation(WSD) is a difficult issue in many fields of natural language processing,e.g.machine translation,information retrieval and hyertext navigation,context and thematic analysis,etc.This paper firstly introduces a Chinese WSD model which is combining the BP neural networks and statistics method,and then discusses the feasibility and advantage of this WSD model.At the last,it finds that the error between the actual and predictive gather is fluctuant through the experiment,namely,the experiment error does not have the notable trend to zero by increasing iterative times,and it can get the better results from the less iterative times.Accordingly,the BP neural network model has the good application foreground in WSD.