基于CNN-BiLSTM网络的电子信息文本快速分类研究
A Study on Fast Classification of Electronic Information Text Based on CNN-BiLSTM Networks
李岷1
作者信息
- 1. 金隆铜业有限公司,安徽铜陵 244021
- 折叠
摘要
传统的文本分类方法往往在处理大规模、复杂的电子信息文本时处理速度慢,导致文本分类的效果差.针对上述问题,提出基于CNN-BiLSTM网络的电子信息文本快速分类研究.首先,计算电子信息文本的相似度,可以通过比较文本之间的语义相似度来实现.接着,获取文本向量的特征权重,通过使用自然语言处理技术来提取文本中的关键词和短语,并计算它们的重要性.然后,CNN提取文本中的局部特征,BiLSTM捕捉文本中的长距离依赖关系.通过将这两种网络结合起来,更好地提取文本的特征向量.最后,判断文本到向量中心的最小距离,从而快速分类电子信息文本.实验证明,该方法能够快速处理大规模、复杂的电子信息文本,文本分类效果好.
Abstract
Traditional text categorization methods often have slow processing speed when dealing with large-scale and complex electronic information texts,resulting in poor text categorization.Aiming at the above problems,a research on fast classification of electronic information text based on CNN-BiLSTM network is proposed.First,the similarity of electronic information texts is calculated,which can be realized by comparing the semantic similarity between texts.Next,the feature weights of the text vectors are obtained,and the keywords and phrases in the text are extracted and their importance is calculated by using natural language processing techniques.Then,CNN extracts the local features in the text and BiLSTM captures the long distance dependencies in the text.By combining these two networks,the feature vectors of the text are better extracted.Finally,the minimum distance from the text to the center of the vectors is judged to quickly classify electronic information text.The experiment proves that the method can quickly deal with large-scale and complex electronic information text,and the text classification effect is good.
关键词
CNN/BiLSTM网络/电子信息/文本快速分类Key words
CNN/BiLSTM network/electronic information/fast text classification引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据