Abstract

In this paper we discuss open information extraction from natural language texts. We present an approach to extraction of semantic relationships using unsupervised machine learning. The presented approach is based on deep clustering methods in which the clusterization algorithm is integrated in a multi-layer auto-encoder neural network. This method allows one to generalize surface relationships (triplets) into semantic relationships. This paper also provides a method of surface relationship extraction.

中文翻译：

---

从文本中提取公开信息：第二部分。使用无监督机器学习提取语义关系

摘要

在本文中，我们讨论了从自然语言文本中提取开放信息的方法。我们提出了一种使用无监督机器学习来提取语义关系的方法。提出的方法基于深度聚类方法，其中将聚类算法集成在多层自动编码器神经网络中。这种方法允许将表面关系（三元组）概括为语义关系。本文还提供了一种表面关系提取方法。