Key points are not available for this paper at this time.
Machine techniques for reducing technical documents to their essential discriminating indices are investigated. Human scanning patterns in selecting topic sentences and phrases composed of nouns and modifiers were simulated by computer program. The amount of condensation resulting from each method and the relative uniformity in indices are examined. It is shown that the coordinated index provided by the phrase is the more meaningful and discriminating.
P. B. Baxendale (Wed,) studied this question.