Boolean Search Now Available
Jon Awbrey
jawbrey at oakland.edu
Sat Apr 19 18:18:37 CEST 2003
o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o
could you tell me more about the primitives that
you have available for the boolean search task?
i was assuming the following sort of framework:
1. you have a population X = {index entries}.
2. x in X has content fields x_1, ..., x_m,
or what amounts to a set of projections,
x_i : X -> X_i = {i^th field entries}.
3. you have a set of computable predicates p : X -> B
that are computable by way of string search operations,
things like "the i^th field of x contains "such & such".
4. for each k you have the boolean functions f : B^k -> B.
5. for each boolean function f : B^k -> B and each k-tuple p
of computable predicates p = <p_1, ..., p_k> : X -> B^k,
you have a boolean query q : X -> B given by this diagram:
q
X o------>o B
\ ^
p \ / f
v /
o
B^k
6. you want to get the models of q in X,
or the "fiber of truth" in X under q.
are those safe enough assumptions so far?
jon awbrey
o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o
N. J. A. Sloane wrote:
>
> One of the troubles with the "lookup" page has always
> been that you could only look up one word or phrase.
>
> So if you wanted to know which sequences mentioned
> "Chess" in the subject line, and "Elkies" in the
> reference line, you had to make two separate searches
> and combine the results by hand.
>
> There is now a new "Boolean" search, that will let you
> specify TWO words or phrases!
>
> You can search for
> P AND Q, P OR Q, or P BUT NOT Q
>
> The program is ready for testing by friendly users.
>
> At present it is in a new advance search page,
> http://www.research.att.com/~njas/sequences/index3.html
>
> When all the bugs are out it will be merged in with
> the regular advanced search page,
> http://www.research.att.com/~njas/sequences/index2.html
>
> Comments are welcomed.
>
> Some things you can do with it:
>
> Find sequences that mention both A012345 AND A001105
>
> Sequences with keyword "uned" AND XXXXX
>
> Sequences that mention "knight" OR "chess"
>
> Sequences that mention "code" (but) NOT "Hamming"
>
> NJAS
o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o
More information about the SeqFan
mailing list