计算机工程与应用 ›› 2019, Vol. 55 ›› Issue (3): 96-103.DOI: 10.3778/j.issn.1002-8331.1807-0228

• 模式识别与人工智能 • 上一篇    下一篇

基于可信语义深度学习的文本文献搜索方法

谢英杰1,2,曾国荪1   

  1. 1.同济大学 计算机科学及技术系,上海 200092
    2.嵌入式系统与服务计算教育部重点实验室,上海 200092
  • 出版日期:2019-02-01 发布日期:2019-01-24

Text Search Method Based on Trust Semantics Deep Learning Neural Network

XIE Yingjie1,2, ZENG Guosun1   

  1. 1.Department of Computer Science and Technology, Tongji University, Shanghai 200092, China
    2.Embedded System and Service Computing Key Lab of Ministry of Education, Shanghai 200092, China
  • Online:2019-02-01 Published:2019-01-24

摘要: 针对传统文本搜索返回结果不准确、不满意的问题,提出一种基于可信语义深度学习的文本搜索方法。首先为了充分挖掘文本的可信语义,通过文本中的信任事实,以及人机交互标注的方式计算文本的可信度。利用网络爬虫抓取大量文本文献学习训练数据,并且构建深度学习神经网络模型,以文本的语义矩阵为输入,以文本的可信度为输出,通过有监督学习,训练出评估文本可信度的深度学习神经网络模型。最后应用该神经网络模型实现文本文献的搜索。通过“中国政治党建”领域的搜索实验表明:该方法在平均可信度方面优于传统搜索方法。

关键词: 文本文献, 信任事实, 可信语义, 深度学习, 智能搜索

Abstract: Aiming at the inaccuracy and dissatisfaction of traditional text search results, a text search method based on trust semantics and deep learning is proposed in this paper. Firstly, in order to fully extract the trust worthiness of the text, the trust degree of the text is calculated by means of the trust facts in the text and the annotation of the human-machine interaction: crawling a large amount of text literatures as the learning and training data by a web crawler, at the same time, building a deep learning neural network model, and training this model with supervised learning for evaluating the trust degree of text, by using the semantic matrix of text as input and the trust degree of text as output. Finally, the neural network model is applied to text and literature search. Through the search experiments in the field of “Chinese politics and party building”, it is shown that the proposed method is superior to the traditional search in terms of average trust degree.

Key words: text and literature, trust fact, trust semantics, deep learning, intelligent search