Chinese patent text similarity calculation method

Abstract

The invention relates to a Chinese patent text similarity calculation method, which is used for calculating text similarity based on the fusion of an SAO structure and a vector space model, and comprises the following steps: extracting an SAO triple from a patent text; calculating the similarity of words in the SAO triple; calculating the similarity between the SAOs; calculating the patent text similarity based on the SAO; and fusing the vector space model method and the SAO structure-based method. According to the Chinese patent text similarity calculation method provided by the invention, the patent text similarity is calculated on the basis of the SAO structure and vector space model fusion method, the advantages of the SAO structure and the vector space model are brought into full play, the defects of the SAO structure and the vector space model are overcome, excellent correct rate, recall rate and F value are obtained, and the requirements of practical application can be well met.

Description

technical field [0001] The invention belongs to the technical field of text processing, and in particular relates to a method for calculating the similarity of Chinese patent texts. Background technique [0002] Patent documents are the carriers of technology, and more than 90% of the technologies in the world are preserved in the form of patent documents. With the explosive growth of knowledge, patent plagiarism and plagiarism are also further increasing. In order to maintain the legitimate intellectual property rights of individuals, enterprises or patent holders can conduct patent lawsuits, patent invalidation applications and patent infringement judgments. Among them, a very important work is to search for related or similar patents in the patent database. However, in the face of the massive patent literature library, the traditional method is to simply enter keywords in the search box to search for related patents. Although there are many results obtained by this metho...

AI Technical Summary

Problems solved by technology

[0004] In the study of text similarity, traditional string-based methods only consider the matching or co-occurrence of strings literally, without taking into account the semantic information contained in the text; ontology-based methods are limited by human-constructed Due to the scale of the ontology, the similarity cannot be calculated for words that are not in the ontology; while the corpus-based method trains the word vector through the neural network and expresses the sentence as a vector form, which can capture the grammatical and semantic information in the text to a certain extent, but the The method does not take into account the positional relationship of the words in the sentence

The method based on the SAO structure takes into account the order of the subject-verb-object in the text, but cannot extract more keywords in the sentence

Images