Small sample patent classification method based on feature map construction

Abstract

The invention discloses a small sample patent classification method based on feature map construction. The method comprises the following steps: S1, obtaining text data of patent documents; S2, extracting a keyword vector in a text form of each patent document, and constructing vocabularies; S3, according to the vocabulary, constructing a patent feature map of the plurality of patent literature; and S4, constructing a neural network model, training the neural network model through the patent feature map, and completing construction of a patent classification model. According to the patent classification method, vocabularies are firstly constructed after keyword vectors are obtained; according to the vocabulary, constructing a patent feature graph of the plurality of patent literature; and constructing a training set pair through the patent feature map to train the neural network model, thereby realizing construction of a patent classification model. According to the patent classification method, a large number of patent literature data sets of the industry are not needed, and meanwhile the classification accuracy reaches 80%; the invention has a good effect in small sample patent classification.

Description

technical field [0001] The invention relates to the technical field of patent analysis, in particular to a small-sample patent classification method based on feature map construction. Background technique [0002] Patent documents have great research value. If the patent documents can be accurately analyzed, important technical details and relationships will be revealed, and this will explain the trend of technology and business development. Novel industrial solutions will also be inspired and proposed, thus Make critical investment decisions. Therefore, patent documents must be carefully analyzed to take advantage of the value of patents. [0003] Existing patent documents are huge, and the large number of patents brings even greater challenges to patent review, management, and analysis and mining. [0004] For example, patent analysis is a very heavy task, which requires a lot of manpower. In general, patent analysts require some level of expertise in a variety of diffe...

AI Technical Summary

Problems solved by technology

[0009] Ma Shuanggang pointed out in "Research on Automatic Classification of Chinese Patent Documents Based on Deep Learning Theory and Method" (Jiangsu University, 2016) that with the development of computer science, the ability of computers to process information has been greatly enhanced, which provides a basis for the automatic classification of patent documents. and designed an automatic classification method for Chinese patents based on deep learning theory. Compared with the classical algorithm, it has achieved better classification results. The disadvantage of this method is that it needs a large amount of patent literature data sets to improve the classification accuracy.

The number of patent documents in our country is increasing day by day, and the standards for the classification of patent documents are also more detailed. However, when the existing automatic patent classification methods are applied to specific industries, the classification effect is not good. The patent literature dataset of

Images