Abstract
Patent citations are significant components of patents, which play a vital role in the implementation of patent analysis. However, most of the existed models only focus on the text of patents and do not realize that citations can remedy missing information in the text. A method for citation modeling in patent analysis is proposed to generate patent citation trees in this paper. Correspondingly, a specific neural network is designed for extracting abstract features in patent citation trees. Then, on the basis of extracted features, a new citation-based vector space model (CVSM) combining citations with text of the patent database is constructed for the subsequent applications. An experiment is conducted based on real patents of USPTO. The experimental results show that the proposed CVSM has good performances in several applications, which demonstrate the effectiveness of the proposed CVSM.
Highlights
The number of patents is growing rapidly with the development of science and technology
With the development of natural language processing (NLP), some traditional NLP algorithms were applied in patent analysis to mine patent text potential information. [24] extracted the significant and rare keywords by Term Frequency-Inverse Document Frequency (TF-IDF) from patent text. [25] used dependency relationships to perform semantic analysis. [26] applied pre-trained Latent Dirichlet Allocation (LDA) models and dependency trees to conduct patents prior-art search. [27] studied the LDA algorithm results for different classes of patents. [28] extracted the descriptions of sci-tech effects and morphological features based on TF-IDF and links between words
APPLICATIONS OF citation-based vector space model (CVSM) In order to validate the effectiveness of citation trees convolution neural networks (CCNN) on patent analysis, various experiments are conducted based on the data set composed of real patents
Summary
The number of patents is growing rapidly with the development of science and technology. It should be noted that there are various types of data provided by a patent besides patent text, which can be divided into two parts, i.e., unstructured data (such as title, abstract, description, claims) and structured data (such as filed dates, inventors’ names, International Patent Classification (IPC) codes, citations) [19], [20] By analyzing such comprehensive data, useful features can be abstracted for developing a more effective model. Citations can tell researchers what areas of technologies the claimed invention might involve [21] Such citation data reflects the relationship among the related patents and is very helpful for patent analysis by clustering the patents and understanding the history of given patents [22], [23].
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
