The Research Of Intelligent Extraction To Information Of High-Yield Management Techniques For Moso Bamboo Stands

Moso bamboo is an important bamboo species with a long cultivation history and the widest distribution area in China.It plays a key role in carbon sequestration and increasing sink,mitigating climate change,preventing soil and water loss,maintaining biodiversity,providing high-quality processed building materials and green and healthy food.The high-yield information of Moso bamboo forest refers to the main indicators such as area,column density,bamboo volume,average bamboo height and average DBH.It is the basic data of the industrial chain of Moso bamboo cultivation,cutting,processing and utilization.For a long time,although all localities and departments have investigated and counted the data related to the high-yield of Moso bamboo,due to the lack of information exchange,a unified database has not been formed.This paper studies the intelligent extraction method of high-yield information of Moso bamboo Forest based on natural language processing,hoping to provide technical support for solving the problems that the data related to high-yield of Moso bamboo are scattered and difficult to find and use.The main research contents and results of this paper include:1)Study on text conversion of data related to high-yield of Moso Bamboo.Use web crawlers and other means to obtain publicly published literature,survey reports and national statistical data from the Internet.Based on the analysis of literature content,PDF and docx documents are read respectively by using PDFBox and POI framework based on Java language,and then the collected literature,reports and statistical data are converted into text format according to the writing program of document structure and content.Experiments show that the proposed method can better convert the mainstream document format codes such as PDF and docx into text format,and maintain the correctness and integrity of its content,and its conversion effect has met the needs of subsequent processing.2)Research on high-yield data extraction of Moso Bamboo Based on deep learning.Through the analysis of the converted text content,it has the problems of small number of target samples and difficult to construct corpus.So a data extraction technology based on "Word classification" and BERT model is proposed,which is suitable for the model training method of target information with small number of target samples and difficult to construct corpus.Experiments show that this method can accurately obtain the target data from the target text,the accuracy rate is higher than the traditional data acquisition method based on text matching,and has a certain error target recognition ability.3)Development of intelligent extraction system for high-yield information of Moso bamboo forest.The high-yield database of Moso bamboo forest and its corresponding web-based access system based on Java and Python are designed,which integrates the above text conversion and automatic extraction modules to realize the batch processing of high-yield data of Moso bamboo.