结合特征降维技术的语言机器学习系统研究  

Research on Language Machine Learning System Combined with Feature Dimension Reduction Technology

作　　者：孙静[1] 况灵巧 SUN Jing;KUANG Lingqiao(Yan’an University Xi’an Innovation College,Xi’an 710100,China)

机构地区：[1]延安大学西安创新学院,西安710100

出　　处：《自动化与仪器仪表》2025年第3期135-139,共5页Automation & Instrumentation

基　　金：第十一批“中国外语教育基金”项目《POA理论指导下的混合式大学英语课程思政体系构建与实践》(ZGWYJYJJ11A031)。

摘　　要：针对智能家居系统在英语指令识别方面效率较低、设备数量增加导致英语指令识别冲突较高的问题,提出了一种结合特征降维技术的英语语言机器学习系统。首先结合自然语言处理和机器学习相关技术进行了系统的设计与分析,其次,利用线性循环网络、多头注意力机制和间隔卷积提出了一种时空多特征提取为基础的语言挖掘算法。实验验证显示,研究所提算法比其他算法的准确率平均增加了5.19%,平均精度平均增加了6.80%。所提算法基础上的系统检测准确率高达99.43%。结果表明,时空多特征提取基础上的语言挖掘算法能够提高智能家居系统对英语指令的识别检测精度,提高对文本信息特征的提取效率。研究提出的系统在智能家居系统指令检测和识别领域具有积极的应用价值。Aiming at the problems of low efficiency of smart home system in English command recognition and high conflict of English command recognition due to the increase in the number of devices,an English language machine learning system combined with feature dimensionality reduction technology is proposed.Firstly,the system is designed and analysed by combining natural language processing and machine learning related techniques.Secondly,a spatio-temporal multi-feature extraction based language mining algorithm is proposed using simple recurrent unit,multi-head attention and dilated convolution.Experimental validation shows that the accuracy of the proposed algorithm under study has increased by an average of 5.19% over other algorithms and the average precision has increased by an average of 6.80%.The accuracy of detection of the system based on the proposed algorithm is as high as 99.43%.The results show that the language mining algorithm based on spatio-temporal multi-feature extraction is able to improve the accuracy of recognition and detection of English commands in the smart home system and improve the efficiency of extracting features of textual information.The system proposed in the study has positive application value in the field of instruction detection and recognition for smart home systems.

关 键 词：特征降维 NLP 英语指令识别 SHS 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]