Multilevel feature model for performing information extraction across fields and feature evaluation method

Abstract

The invention relates to a multilevel feature model for performing information extraction across fields and a feature evaluation method. According to the method, features in existing information extraction relevant literature are utilized to construct an original feature library; a multilevel feature theoretical model is constructed, the features of specific fields are degraded, the features are divided into composite features and atomic features, and field relevance of the features is lowered; a feature adaptive evaluation method is proposed based on the multilevel feature theoretical model, a sample library is used to evaluate cross-field adaptability of the features capable of being obtained, and the features which can be used quickly and repeatedly are obtained; and the ability of the multilevel feature theoretical model in adapting to field changes is utilized to perform changeability management on the features in the model, actual webpage analysis and processing and feature recognition, matching and parameterization in the actual webpage, and a webpage information extraction feature evaluation system is realized. Through the multilevel feature model and the feature evaluation method, modeling of multilevel features in the information extraction feature evaluation system is completed, and a cross-field information extraction function with high adaptability is achieved.

Description

technical field [0001] The invention relates to a multi-level feature model and a feature evaluation method for cross-domain information extraction. Background technique [0002] The information extraction method is a method of extracting information of interest to users from semi-structured and unstructured documents and structuring it. It has been widely used and recognized in Internet content retrieval with rapidly increasing information volume. [0003] The cross-domain problem of information extraction refers to the adaptability of information extraction methods to the information extraction tasks of different subject contents and different forms of documents. The field includes two aspects: on the one hand, it refers to the subject of information, for example, the information extraction model for sports news is difficult to directly apply to the information extraction of travel strategies; on the other hand, it refers to the form of information, for example, for the pr...

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to overcome the deficiencies in the prior art. Based on text features (knowledge topics and text forms), a multi-level feature model and feature evaluation method for cross-domain information extraction are provided to target cross-domain information extraction. problem, establish a multi-level feature model, and improve the domain adaptability of the information extraction method

Images