MONO91
NIST Monograph 91: Automatic Indexing: A State-of-the-Art Report
Other Potentially Related Research
chapter
Mary Elizabeth Stevens
National Bureau of Standards
a posteriori associations are at least foreshadowed in a small-scale model of attribute-
words and proper names, together with prespecified relationships between them; 1/ in
Olney's recent work at SDC exploring the possibilities for use of cognitive concepts as
bases for establishing association between documents, 2/and by Kochen1s work on machine'
inference and concept processing 3/
A final example of potentially related research in the area of content analysis is
therefore the work of Kochen, Abraham, Wong and others at IBM's Thomas J. Watson
Laboratories (1962 [329]). While concerned principally with adaptive organization and
processing of stored factual statements and the possibilities for machine formulation
of 11hypotheses" about these and additional facts, some consideration has been given to
sampling procedures applicable to determination of similarity which might be used for
document clustering and to the possibilities for dynamic clustering for retrieval based
upon a specific individual query. 4/ In the proposed AMNIP (Adaptive Man-machine Non-
Arithmetical Information Processing) system,, there is no attempt at either automatic
indexing or automatic abstracting. 5/ Instead, formal statements are made about named
1'things" and their attributes. The sharing of common attributes then serves as a basis
for relating items which are similar and for grouping them together in the system
memory. It is assumed that the organization of the stored statements changes dynami-
cally with new data inputs and user feedback in question-answering routines.
Where the named items are names of documents or of index terms, a number of
documentation applications can be considered. Where the items are document names
and the formal predicate is "cites'1, the system provides a procedure for production and
use of citation indexes. 6/ Where the items are index terms or subject headings and
the predicates are "i5 used synonymously with" or "is subsumed under'1, machine
construction of a growing thesaurus based on use is suggested. 1/ The common attribute
1/
2/
See Stevens, 1960 [568]; see also Herner, 1962 [266], p. 5.
See Borko, 1962 [75], p. 5: "Instead of defining meaning in terms of synonyms...
it is defined in terms of the entities referred to by the word in context. A chair
is thus described as belonging to a class defined by a given list of properties.
Analysis yields an interpretation of the sentence as an assertion that certain
relationships hold between the specified referent classes. The cognitive content
of the sentence is a function of this assertion plus the information about these
referent classes which has previously been stored in memory."
3/
4/
5/
6/
7/
Kochen et al, 1962 [329]
Ibid, Appendix by C. T. Abraham, pp. 20-65.
Kochen et al, 1962 [328], p. 45.
Ibid, p. 37.
Ibid, p. 37.
139