A Hierarchical Neural Network Document Classifier with Linguistic Feature Selection

摘要

In this article, a neural network document classifier with linguistic feature selection and multi-category output is presented. It consists of a feature selection unit and a hierarchical neural network classification unit. In the feature selection unit, the candidate terms are extracted from some original documents by text processing techniques, and then the conformity and uniformity of each term are analyzed by an entropy function which can measure the significance of terms. Terms with high significance are selected as input features for training neural network document classifiers. In order to reduce the input dimensions, a composition mechanism of fuzzy relation is employed to identify synonyms. By this method, a synonym thesaurus can be constructed to reduce input dimensions. To simplify the learning scheme, the well-known back-propagation learning model is used to build proper hierarchical classification units. In our experiments, a product description database from an electronic commercial company is employed. The experimental results show that this classifier achieves sufficient accuracy to help human classification. It can save much manpower and work time classifying a large database.