The use of title and cited titles as document representation for automatic classification

摘要

In this investigation the use of title and cited title words as document representation is explored. It ofters a method intermediate between the use of title and abstract of a document and that of citation identities, retaining some advantages of both. Compared with title and abstract, it leads to more compact and uniform document representation with a high concentration of indicative words, gives more consistent coupling strengths to profiles with results agreeing well with that employing citations, and offers a more consistent ability for inter-group differentiation when the groups are close to each other. Compared with the use of citations, it gives results with less specificity and operationally requires an extra step to input and analyse the full citation titles. However, the group profiles derived from title and cited titles are words and can be used to classify documents that have descriptive abstracts but no or few citations.