AUTOMATIC EXTRACTION OF CITATIONS FROM THE TEXT OF ENGLISH-LANGUAGE PATENTS - AN EXAMPLE OF TEMPLATE MINING

Methods for automatically isolating and extracting bibliographic refer ences from the full texts of patents are described and evaluated; thes e include citations both to patents and to other bibliographic sources . Patents are unusual as citing documents in that citations occur prin cipally in the text of the abstracts or description parts of tbe docum ents, rather than as footnotes or in separate sections. A template min ing approach has been developed for this purpose, to relieve patent ex aminers of the chore of doing this manually. The sub-languages of cita tions in patents are examined, and the development of templates for th e extraction of citations to patents, journal articles, books and othe r sources in English-language patents described, as well. as the evalu ation of the degree of success of the approach.