面向深度学习的智慧图书馆文本相关性资源数据检索模型设计

Design of Deep Learning Driented Data Retrieval Model for Text-Related Resources in Smart Library

随着图书资源数字化技术的发展,传统的检索方式已经无法满足当前的图书资源海量数据查询的需求。针对智慧图书馆发展需求,本文提出了一种深度学习模型,用于将文档文本与关键字样式查询相关性;通过相对较少的训练数据,模型使用预训练词嵌入,首先计算查询和文档之间的可变长度Delta矩阵,描述两个文本之间的差异,然后将其传递到深度卷积阶段,再经过深度前馈网络以计算相关性得分。形成了适用于在线搜索引擎的快速模型,实验结果证明该模型性能优于同类的最新深度学习方法。

With the development of digital technology of book resources, traditional retrieval methods can no longer meet the current demand for massive data query of book resources. In response to the development needs of smart libraries, this paper proposes a deep learning model to correlate document text with keyword style queries;with relatively little training data, the model uses pre-trained word embeddings, and first calculates the difference between the query and the document the variable-length Delta matrix between the two texts, it describes the difference between the two texts, and then passes it to the deep convolution stage, and then passes through the deep feedforward network to calculate the correlation score. A fast model suitable for online search engines is formed, and the experimental results prove that the performance of this model is better than the latest deep learning methods of the same kind.