Research Topic Identification And Evolution Analysis In The Field Of Agricultural Engineering Based On LDA-BERT-K-means Mode

At present,with the rapid development of computer technology,text mining technology has made unprecedented progress and has become one of the important methods for research topics in the field of analysis.At the same time,as an important support for the development of national agricultural modernization,agricultural engineering technology has many research results in its field and lacks systematic research theme analysis,so paying close attention to the distribution and evolution trend of research topics in the field of agricultural engineering is of great significance to accelerate the process of national agricultural modernization and promote agricultural scientific and technological innovation.This paper takes high-level journal papers at home and abroad in the field of agricultural engineering as the research object,first collects relevant literature at home and abroad in the past ten years as the original data,divides the domestic and foreign data into two stages according to the time average,and carries out data cleaning and data structuring to construct a literature data corpus in the field of agricultural engineering.Secondly,the optimal number of topics is determined by calculating the perplexity,and the traditional LDA theme model is used to analyze the research topics in the field of agricultural engineering.Finally,based on the shortcomings of LDA topic model and combing the mainstream models in text mining,a new text topic recognition method based on LDA-BERT-K-means model is proposed.Firstly,the model makes up for the lack of contextual semantic association of the traditional LDA theme model by splicing LDA thematic feature vectors and BERT semantic feature vectors to a certain extent,and secondly,the K-means clustering algorithm is used to perform semantic association clustering analysis on the splicing post-vectors,and the identification of research topics in the field of agricultural engineering is realized by combining two-dimensional clustering results and feature words in different clusters.Finally,the validity of the model is verified by constructing the coherence index of the theme and comparing the calculation results with the traditional theme model.The experimental results show that the LDA-BERT-K-means model proposed in this paper effectively improves the thematic coherence compared with the traditional LDA theme model,and performs better in the two-dimensional visualization of clustering results,and the distribution of research topics in the field of agricultural engineering can be better identified by using the model.In the research texts in the field of domestic agricultural engineering,a total of two stages and eight themes were identified,and the research texts in the field of foreign agricultural engineering identified two stages and seven themes,and after the analysis of the distribution of different themes,it was concluded that the change of research themes in the field of agricultural engineering in China in the past ten years showed a trend of excessive from agricultural mechanization engineering to agricultural modernization engineering.In the past ten years,the field of foreign agricultural engineering has shown a trend of gradual deepening of agricultural modernization related technologies with time in the change of research themes.