Functional Structure Tuple Extraction Based on Statistics and Rules

摘要

When the vocabulary model represents the features of patent document, the documents are divided into words that are not related to each other. It is difficult to extract the semantic features of the document. If the relationship between words in a sentence is taken into consideration, the functional structure tuples, which extracted as feature of the engineering technology documents, can further improve the classification effect. This article starts with the characteristics of patent document, explores the characterization methods of patent document, and proposes relevant ideas for the extraction of functional structure tuples. After verification, this method can effectively improve the feature extraction effect of patent document.