AUTOMATIC ABSTRACTING OF RUSSIAN TEXT.
TRW COMPUTERS CO CANOGA PARK CALIF
Pagination or Media Count:
The report describes the development of a technique that is based on previously tested methods for English text. During the study an operating system was developed which produces automatic abstracts of Russian text by four basic methods Cue, Key, Title, and Location which were programmed for the IBM 7094. The programming involved the preparation of an edit program which prepares an edit tape of the text of Russian documents, a cue dictionary program which prepares a tape of a fixed word list, and an abstracting program which selects and outputs sentences of the Russian document in transliterated form. An example is given of a machine abstract produced by the combined method, using Cue, Title, and Location, which was found to be the preferred method. For purposes of comparison, an abstract the target abstract created manually according to a set of specific instructions is also presented. Also included are the conclusions resulting from the study and recommendations for future research. Author