MONO91
NIST Monograph 91: Automatic Indexing: A State-of-the-Art Report
Other Potentially Related Research
chapter
Mary Elizabeth Stevens
National Bureau of Standards
Hughes provides, as of September, 1962 ([284]), a critical review of several
experimental and proposed question-answering systems using natural language statements
and natural language queries, including "BASEBALL", 1/ "SAD SAM" 2/ and the "Proto
Synthex" investigations of System Development Corporation. Later developments on
the Synthex (synthesis of complex verbal material) project at SDC have included a
variation on a natural language text searching program where ordinary text input is run
against an exclusion list and a table is set up to tally the substantive words remaining.
Words with the same roots or previously having been identified as synonymous are cross-
referenced. A complete index results, with document location identifier tags for the
word occurrences down to the single sentence level. This index can be used subsequently
to locate regions of text (volume, chapter, paragraph, and sentence) where answers
responsive to input questions are likely to be found.
It is proposed that the Synthex system eventually should incorporate analyses of
syntactic and semantic relationships in the linguistic expressions of both queries and
text. Of future interest in the extension of such considerations to automat[OCRerr]c indexing and
abstracting are the following comments:
"The results of several early experiments within the project, coupled w[OCRerr]th the
findings of other language researchers, led to the following conclusions about
meaning and grammatical structure in English text:
1. The degree of synonymity in meaning between any two English
words can be measured quantitatively with a synonym dictionary
and relatively simple scoring procedures.
2. The difference in meaning between two sentences of identical
syntactic structure can be expressed quantitatively as a function
of synonymity of their words..." 4/
It is also of interest to note that although the "indexer" program of the Synthex
system provides cross-referencing between, for example, "whales" and "whaling" or
"England" and "Great Britain", the investigators admit that: "naturally it falls short of
such complicated cross-referencing as `mouse-animal' `Jones person' and other
concept recognitions." 5/ However, concept recognitions based upon both a priori and
1/
2/
3/
4/
5/
See also Green et al, 1961 [238].
See also Lindsay, 1960 [363 ]
See also Klein and Simmons, 1961 [325 j; Simmons et al [552]to [555].
System Development Corporation, 1962 [590].
Simmons and McConlogue, 1962 [ 555 [OCRerr] , p.70.
138