MONO91
NIST Monograph 91: Automatic Indexing: A State-of-the-Art Report
Other Potentially Related Research
chapter
Mary Elizabeth Stevens
National Bureau of Standards
documents relevant to a request even though the documents had not been indexed
by the terms used in the request." 1/
In another case, which was analyzed in detail, a request profile of 26 terms that
had been intuitively weighted by the customer resulted in the machine listing of 246
presumably responsive documents. Of these, 81 documents were of primary interest
to the customer, and an additional 78 were of secondary interest to him. 2/
The statistical association technique as proposed by Stiles has also been investi-
gated at the Datatrol Corporation, with particular reference to the field of legal
literature (Hammond et al, 1962 [251]). About 350 documents in the field of Federal
public law were indexed in cooperation with George Washington University, using a
vocabulary of 680 index terms. A computer program was written for the IBM 7090
can accommodate a 1200 x 1200 matrix to calculate the Stiles' association factors.
were made of various thresholds to determine which other terms were sufficiently
association strength to a particular term to be selected for that term's profile.
that
Trials
high in
Given the generation of the term profiles, a less sophisticated computer such as
the 1401 can be used for the expansion of request terms and the actual conduct of searches.
Such a program was demonstrated at the Annual Meeting of the American Bar Association,
August 1962, with running of "live" requests suggested by jurists and with what are
claimed to be "highly gratifying results". A point of interest relates to the question of
updating of term-profiles and other statistical association factor data. Hammond, et al
report:
"The term profiles were generated a total of three times in the course of the
pilot study, making it possible, to some extent, to assess the effect of
vocabulary growth. Judging from this limited experience, it appears that a bi-
monthly, or perhaps even quarterly, recompilation of term profiles should be
sufficient for a mature collection." 34
6.2.3 The Association Map - Doyle and Related Work at SDC
The name of Doyle is again that of an early and prolific investigator and innovator
in the field of mechanized documentation and linguistic data processing. One of his
provocative suggestions is generally known, in his own terminology, as that of "semantic
road maps for literature searchers" or an "association map" technique. As a matter of
convenience, we have chosen to consider this suggestion and a variety of related work
1/
2/
3/
Stiles, 1961 [577], pp. 198-199.
Stiles, 1962 [573], p.9.
Hammond et al, 1962 [251], p.6.
[OCRerr]22