Combination of improved cosine similarity and patent attribution probability method to judge the attribution of related patents of hydrolysis substrate fabrication process

• Innovatively establishing a combination method of judging patent attribution.
• Improved cosine similarity concept and patent attribution probability is combined.
• Term and word segmentation system is used to analyze technical/functional words.
• Normalized numerical value is used as the weight of improved cosine similarity.
• Normalized numerical value is used in the probability method.

The paper studies the attribution of patents by innovatively establishing a combination of improved cosine similarity concept and patent attribution probability method for hydrolysis substrate fabrication process in order to enhance the speed and accuracy for judgment of patent attribution. For the improved cosine similarity method innovatively established in the paper, all the vocabularies of important technical words or functional words in patent documents are regarded as a number of vector dimensions. The normalized numerical values of these vocabularies are regarded as the weights of these technical words or functional words. They are substituted in a formula of improved cosine similarity. Regarding the patent attribution probability method, it applies the normalized numerical values of the various clusters of technical or functional words as well as the formula of probability method to judge which technical category or functional category that a patent is attributed to. As the study innovatively combines improved cosine similarity with patent attribution probability method, before employing patent attribution probability method for a patent document, the study firstly uses improved cosine similarity to check with the cluster of patent group words having high relativity, and rule out the unrelated technical categories or functional categories that the patent is not attributed to, so as to decrease the number of categories for calculation and the time to calculate which technical category or functional category that a patent is attributed to when using patent attribution probability method, and to achieve a method that can more rapidly and accurately judge which technical category or functional category that a patent is attributed to.