Research On Technology Trend Prediction Method Based On Word Vector

In today's world,the science and technology iterative speed is ever increasing and the acceleration of economic globalization make the countries and enterprises are facing increasingly fierce competition.Therefore,judging the development of a technology and predicting technological trend is very important,countries and enterprises can adjust their research strategy to get a good competitive position in future.At present,most of the prediction methods of technical development trend are subjective patent analysis method,but this method needs very professional experts,and the results are based on people's subjective judgment.Although there are some studies on the use of computer technology to analyze the patent data to predict the results,but few studies can make full use of the semantic features of the patent text.Word2Vec is a kind of distributed representation of lexical semantics,commonly used in the calculation of lexical similarity,this paper will process this method into the patent text mining,combined with the theory of TRIZ,introduces the technology development trend prediction method based on word vector.First of all,the technical feature words are extracted from the text of the patent,and the Word2 Vec of the technical theme is trained,the technical feature words and word vectors are combined to form the feature vector.Secondly,according to the theory of TRIZ in patent classification standard,using Word2Vec's ability to calculate semantic similarity,thinking of the technical feature source and backreference of patent,put forward a patent invention level classification model base on word vector.Finally,the paper uses the classification results of patent level to identify the technology maturity and technology track,so as to achieve the purpose of technical development trend prediction.In this paper,the speech recognition technology is taken as an example.The technology development trend prediction method from this paper is used to predict the future development trend of speech recognition technology,and the feasibility and effectiveness of the proposed method are verified.