基于对抗生成网络的命名实体识别

Named Entity Recognition Based on Adversarial Generative Network

摘要: 随着时间的流逝，互联网技术的迅速发展，如何从大量的文本数据中获取对我们有用的信息成为了一种新的挑战。命名实体识别是信息抽取和信息检索中一项重要的任务，其目的是识别出文本中表示命名实体的成分，并对其进行分类。它在垃圾邮件过滤，舆论分析和邮件分类等许多领域中广泛使用并发挥重要作用。考虑到采用对抗生成网络的形式能够更好地学习到样本数据的特征分布和采用变分自动编码器能够更好地接近真实样本的优点，本文通过对抗网络的形式，将两种现阶段具有各自特点的先进模型进行了结合，综合设计了一种基于对抗生成网络的命名实体识别算法模型用以提高特征提取的准确有效性。

Abstract: With the passage of time and the rapid development of Internet technology, how to obtain useful information from a large amount of text data has become a new challenge. Named entity recognition is an important task in information extraction and information retrieval, and its purpose is to identify and classify the components that represent named entities in the text. It is widely used and plays an important role in many fields such as spam filtering, opinion analysis and message classification. Taking into account the advantages of using adversarial generation networks to better learn the feature distribution of sample data and the advantages of using variational autoencoders to better approximate real samples, this article uses the form of adversarial networks to separate the two current stages. The advanced models with characteristics are combined to design a named entity recognition algorithm model based on adversarial generative network to improve the accuracy and effectiveness of feature extraction.