Some Issues in the Automatic Classification of US Patents,
MASSACHUSETTS UNIV AMHERST DEPT OF COMPUTER SCIENCE
Pagination or Media Count:
The classification of US patents poses some special problems due to the enormous size of the corpus, the size and complex hierarchical structure of the classification system, and the size and structure of patent documents. The representation of the complex structure of documents has not been a standard area of research in text categorization, but we have found it to be an important factor in our previous work on classifying patient medical records Larkey and Croft, 1996 and in our current work on US patents. Our classification approach is to combine the results of k-nearest-neighbor classifiers with those of Bayesian classifiers. The k-nearest-neighbor classifier allows us to represent the document structure using the query operators in the Inquery information retrieval system. The Bayesian classifiers can use the hierarchical relations among patent subclasses to select closely related negative examples to train more discriminating classifiers.
- Information Science