北京大学学报自然科学版 ›› 2019, Vol. 55 ›› Issue (1): 22-28.DOI: 10.13209/j.0479-8023.2018.061

上一篇    下一篇

一种基于Hownet的词向量表示方法

陈洋1, 罗智勇2,†   

  1. 1. 北京语言大学信息科学学院, 北京 100083
    2. 北京语言大学语言信息处理研究所, 北京 100083
  • 收稿日期:2018-04-15 修回日期:2018-08-18 出版日期:2019-01-20 发布日期:2019-01-20
  • 通讯作者: 罗智勇, E-mail: luo_zy(at)blcu.edu.cn

A Word Representation Method Based on Hownet

CHEN Yang1, LUO Zhiyong2,†   

  1. 1. College of Information Science, Beijing Language and Culture University, Beijing 100083
    2. Institute of Linguistic Information Processing, Beijing Language and Culture University, Beijing 100083
  • Received:2018-04-15 Revised:2018-08-18 Online:2019-01-20 Published:2019-01-20
  • Contact: LUO Zhiyong, E-mail: luo_zy(at)blcu.edu.cn

摘要:

针对基于预训练得到的词向量在低频词语表示质量和稳定性等方面存在的缺陷, 提出一种基于Hownet的词向量表示方法(H-WRL)。首先, 基于义原独立性假设, 将 Hownet中所有N个义原指定为欧式空间的一个标准正交基, 实现Hownet义原向量初始化; 然后, 根据Hownet中词语与义原之间的定义关系, 将词语向量表示视为相关义原所张成的子空间中的投影, 并提出学习词向量表示的深度神经网络模型。实验表明, 基于Hownet的词向量表示在词相似度计算和词义消歧两项标准评测任务中均取得很好的效果。

关键词: 词向量表示, Hownet, 词语相似性计算, 词义消岐

Abstract:

Word embedding method based on pre-training still has some defects in the stability and the quality of low-frequency words. The authors propose a new word embedding method based on Hownet. First, based on the sememe independence assumption, all sememes of Hownet are specified in an Euclidean Space’s standard orthogonal basis to initialize all sememe vectors. Secondly, utilizing the relationship between word and sememe defined in the Hownet, each word vector representation can be regarded as a subspace projection by related sememes. Finally, a deep neural network model is put forward to learn word representations. The experimental results indicate that proposed word embedding method based on Hownet obtained comparable results in the two standard evaluation tasks including the word similarity computation and the word sense disambiguation.

Key words: word embedding, Hownet, word similarity computation, word sense disambiguation