Patent value information analysis method and device based on multi-semantic fusion

Abstract

The invention provides a patent value information analysis method and device based on multi-semantic fusion, and the method comprises the following steps: obtaining the numerical information, text information and image information of a patent from a patent database, and carrying out the preprocessing; inputting the text information test set into a trained semantic classification model Bert, and extracting text semantic features; inputting the image information test set into a trained image classification model DenseNet, and extracting image features; inputting the preprocessed numerical information into an XGBoost classifier as an unstructured data feature, inputting a text semantic feature into an MLP classifier, inputting an image feature into a ResNet classifier, and respectively obtaining classification results; and analyzing patent values according to classification results of the three classifiers. According to the method provided by the invention, multiple features such as unstructured data features, text semantic features and image features of patents are extracted and analyzed by applying a leading-edge deep learning technology, comprehensive patent value evaluation is achieved, the whole analysis process does not depend on expert experience, the speed is high, and the accuracy is high.

Description

technical field [0001] The present invention relates to the field of patent value analysis, in particular to a patent value information analysis method and device based on multiple semantic fusion. Background technique [0002] At present, the analysis of patent value is basically carried out through the evaluation method of experts in related fields who have been working for many years. This evaluation method largely relies on expert experience and is highly subjective, which brings great risks to the analysis of patent value. And this process is slow, inefficient and expensive. There is a lack of systematic and public-oriented effective patent value analysis methods in the prior art. Contents of the invention [0003] The main purpose of the present invention is to overcome the above-mentioned defects in the prior art, and propose a patent value information analysis method based on multiple semantic fusion. And image features and other multiple features are extracted...

AI Technical Summary

Problems solved by technology

[0002] At present, the analysis of patent value is basically carried out through the evaluation method of experts in related fields who have been working for many years. This evaluation method largely relies on expert experience and is highly subjective, which brings great risks to the analysis of patent value. And this process is slow, inefficient, and expensive

There is a lack of systematic and public-oriented effective patent value analysis methods in the prior art

Images