The Vectorization Method Of Geological Map Based On NLP Technology And Its Application In Prospecting

The Zhaishang area is located in the West Qinling polymetallic metallogenic belt.There are Zhaishang Gold Mine,Mawu Gold Mine,Suolong Gold Mine,Xinzhuangli Gold Mine,Xuehuashan Tungsten Mine,Bangou Lead-Zinc Mine and other deposits.Predecessors have carried out a lot of in-depth studies on the geological characteristics,metallogenic laws,metallogenic models,geophysical and geochemical prospecting characteristics,and the genesis of the ore deposits in the Zhaishang-Mawu ore concentration area.However,the progress of traditional prospecting prediction methods has slowed down,and a new approach is urgently needed.The method brings new progress and new ideas to the prospecting and forecasting work in this area.The application of artificial intelligence and NLP technology in this area is rarely carried out.How to make full use of geological maps and text data,extract the prospecting prediction information,and improve the effect of prospecting prediction is a research direction that needs to be tackled urgently at present.Based on the collection and sorting of maps and data in the Zhaishang-Mawu area,this paper collects text data for the Zhaishang-Mawu mining area,and builds a corpus,trains and acquires four language models and performs word embedding and sentence Embedding,taking the vectorization of the geological map of the Zhaishang mining area as an input layer,using convolutional neural network to predict the prospecting of the Zhaishang mining area,and delineating 4 prospecting target areas.The research work has exploratory significance for the artificial intelligence prospecting prediction method,and at the same time has practical application value for the prospecting and exploration in the Zhaishang-Mawu ore concentration area.The main research contents and knowledge gained include:(1)Improved a set of corpus construction process:collect journal documents,dissertations,reports,monographs and other literature materials for keywords such as"West Qinling" and "Zhaishang-Mawu",constructing more than three million words Geological Corpus.(2)On the basis of the constructed corpus,train language models such as Word2vec and BERT.According to the geological map of the study area,the geological body attribute sentence containing the information of "lithology+age" is sorted out.Input the language model to obtain the sentence vector.Assign sentence vectors to corresponding attribute grids,construct vectorized geological maps,and assign semantic information to geological maps.(3)Use a convolutional neural network to give a grid of attribute text sentence vectors and geophysical and geochemical data as input to the network.Extract the coordinates of known mineral deposits(points)in the study area as the network output to train the network.To predict the prospecting target area.Combined with geological information,a total of 3 prospecting target areas have been delineated in the area.In addition,test the impact of data sets,parameters,word embedding methods,etc.on the prediction effect.It provides a new direction for further prospecting and exploration in the research area.