基于深度学习的领域实体属性词聚类抽取研究  

Study on word clusting for attribute extraction based on deep learning

作　　者：苏丰龙 谢庆华[2] 邱继远[1] 岳振军[1] 

机构地区：[1]解放军理工大学通信工程学院,江苏南京210007 [2]解放军理工大学国防工程学院,江苏南京210007

出　　处：《微型机与应用》2016年第1期53-55,59,共4页Microcomputer & Its Applications

基　　金：江苏省自然科学基金资助项目(BK2012511)

摘　　要：属性词的聚类是领域实体属性抽取中的一个重要步骤。在未知领域和大量文本中,人工标注寻找十分困难。本文将一种基于深度学习框架的词语嵌入表示方法 (Word Embedding)引入到领域实体属性词聚类研究中,在无监督条件下解决大规模语料、领域实体属性词表人工参与构建代价较高的问题,并进行了适当的扩展,取得了较好的效果,可以为信息抽取等后续高级任务提供较好服务。Attributes word clustering is an important step for entity attributes extraction in domain. It is very difficult for man to find them out in a large number of texts about unknown domain. We introduced Word Embedding which is based on a deep learning frame- work into the study of attributes word clustering, in that we should solve the problem of the higher cost of building the attibutes words lists in large scale corpus. Then we brought them to develop more vocabularies and expressions, and the results was good. It can pro- vide us better service in some senior tasks in the follow-up work, such as information extraction.

关 键 词：词向量 深度学习 词语聚类 属性抽取 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]