当前位置: X-MOL 学术ACM Trans. Asian Low Resour. Lang. Inf. Process. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
A Framework for Extractive Text Summarization Based on Deep Learning Modified Neural Network Classifier
ACM Transactions on Asian and Low-Resource Language Information Processing ( IF 2 ) Pub Date : 2020-07-07 , DOI: 10.1145/3392048
BalaAnand Muthu , Sivaparthipan CB , Priyan Malarvizhi Kumar , Seifedine Nimer Kadry , Ching-Hsien Hsu , Oscar Sanjuan , Ruben Gonzalez Crespo

There is an exponential growth of text data over the internet, and it is expected to gain significant growth and attention in the coming years. Extracting meaningful insights from text data is crucially important as it offers value-added solutions to business organizations and end-users. Automatic text summarization (ATS) automates text summarization by reducing the initial size of the text without the loss of key information elements. In this article, we propose a novel text summarization algorithm for documents using Deep Learning Modifier Neural Network (DLMNN) classifier. It generates an informative summary of the documents based on the entropy values. The proposed DLMNN framework comprises six phases. In the initial phase, the input document is pre-processed. Subsequently, the features are extracted using pre-processed data. Next, the most appropriate features are selected using the improved fruit fly optimization algorithm (IFFOA). The entropy value for every chosen feature is computed. These values are then classified into two classes, (a) highest entropy values and (b) lowest entropy values. Finally, the class that holds the highest entropy values is chosen, representing the informative sentences that form the last summary. The results observed from the experiment indicate that the DLMNN classifier gives 81.56, 91.21, and 83.53 of sensitivity, accuracy, specificity, precision, and f-measure. Whereas the existing schemes such as ANN relatively provide lesser value in contrast to DLMNN.

中文翻译:

一种基于深度学习修正神经网络分类器的文本提取框架

互联网上的文本数据呈指数级增长,预计在未来几年将获得显着增长和关注。从文本数据中提取有意义的见解至关重要,因为它为企业组织和最终用户提供了增值解决方案。自动文本摘要 (ATS) 通过减小文本的初始大小来自动进行文本摘要,而不会丢失关键信息元素。在本文中,我们提出了一种使用深度学习修正神经网络 (DLMNN) 分类器的文档文本摘要算法。它根据熵值生成文档的信息摘要。提出的 DLMNN 框架包括六个阶段。在初始阶段,对输入文档进行预处理。随后,使用预处理数据提取特征。下一个,使用改进的果蝇优化算法 (IFFOA) 选择最合适的特征。计算每个所选特征的熵值。然后将这些值分为两类,(a)最高熵值和(b)最低熵值。最后,选择具有最高熵值的类,代表形成最后一个摘要的信息句。从实验中观察到的结果表明,DLMNN 分类器给出了 81.56、91.21 和 83.53 的灵敏度、准确度、特异性、精确度和 f 测量值。而与 DLMNN 相比,现有的方案(如 ANN)提供的价值相对较小。然后将这些值分为两类,(a)最高熵值和(b)最低熵值。最后,选择具有最高熵值的类,代表形成最后一个摘要的信息句。从实验中观察到的结果表明,DLMNN 分类器给出了 81.56、91.21 和 83.53 的灵敏度、准确度、特异性、精确度和 f 测量值。而与 DLMNN 相比,现有的方案(如 ANN)提供的价值相对较小。然后将这些值分为两类,(a)最高熵值和(b)最低熵值。最后,选择具有最高熵值的类,代表形成最后一个摘要的信息句。从实验中观察到的结果表明,DLMNN 分类器给出了 81.56、91.21 和 83.53 的灵敏度、准确度、特异性、精确度和 f 测量值。而与 DLMNN 相比,现有的方案(如 ANN)提供的价值相对较小。精度和 f 测量。而与 DLMNN 相比,现有的方案(如 ANN)提供的价值相对较小。精度和 f 测量。而与 DLMNN 相比,现有的方案(如 ANN)提供的价值相对较小。
更新日期:2020-07-07
down
wechat
bug