[dba-Tech] Questions about 2 Unusual Databases

Arthur Fuller artful at rogers.com
Wed Oct 27 19:46:56 CDT 2004


Thanks for responding! Doubly interesting.... On the first case, part of 
the reason I keep coming back to this is that I have a gut feeling that 
brute force is all wrong for this application. What I have in mind is a 
sort of double-perspective on any given chess situation -- one that 
records the sequence and the other that records the absolute position, 
so that any two positions could be compared very rapidly without having 
to go through the move-sequences to build it. Not to suggest that the 
following is anywhere near an optimal model for recording the latter 
piece, but let's just say it might look something like this....
There are 64 squares and at most two pairs of 16 pieces. (Convenient 
numbers from a computing viewpoint.) So we could have a 1-D array of 64 
elements or a 2-D array of 8*8 elements to reprsent the squares. 
Regarding the pieces, we need to distinguish white from black, but we do 
not need to distinguish Queen's knight from King's. The front row (at 
setup) is an array of 8 pawns; the back row is a ragged array of 3 pairs 
(rook, knight, bishop) and perhaps another pair or two single-element 
items, Q and K.
To record any given position, we need to note the square of interest, 
the piece that's sitting on it and the colour of said piece. If we could 
map this compactly and effectively, we could also search it rapidly, I 
think. Let's say for the sake of argument that positions P1 and P2 
differ by only one piece's position. Let's further say that we have 
employed a legion of low-wage workers to plug in the Book of Endings. 
Then (and here comes a large leap of faith) any position P3 could be 
compared to any known and similar position P4 that is guaranteed to 
result in victory (or defeat). I.e., P3 can be compared with P4 
(victory) and the relatively small problem of how to get from here to 
there can be concentrated upon. If I can paint you into said corner, 
then I'm guaranteed to win and the rest is rote.

----------------
On problem two, I guess that we have both invested some time in this 
investigation, and that's (for me at least) a good thing. I tackled it 
in various ways, from studying and playing music to taking various 
academic courses and reading the literature on various investigations 
from researchers (not to say I'm in any way expert, but I have read 
some). From I gather, the most accurate vector of prediction is what you 
have previously listened to. As it happens, I am either "eclectic" or 
musically promiscuous -- you choose. I have almost everything Beethoven 
wrote, and Bach, and also Cecil Taylor, John Coltrane, Talking Heads, 
the Clash and Zakir Hussain, to name only a few. This opens me to 
radical new musics in a way that is simply unavailable to people 
acquainted with only one or two genres and only four or five decases of 
same. I'm not trying to toot my horn here, and if the previous sentence 
reads that way then I apologize. Here's another perspective on the point 
I'm attempting to make: when I started studying classical guitar about 
-- god! -- 35 years ago, it took me on average about 20 minutes to tune 
the guitar. Once two notes got very close to each other, I had a very 
difficult time determining which one was higher than the other. A gifted 
friend of mine cleared up the mud with a simple instruction: listen to 
the wa-wa-wa as you compare the notes. The faster the wa-wa-wa, the 
further the notes are apart. Adjust the pegs and make the wa-wa-wa 
slower. Once it gets to a "wa" per second, you're close enough for folk 
music. After he told me that, my time to tune shrank dramatically, and 
now I can sit in the back row of a nightclub and tell you in seconds 
who's out of tune.
That doesn't mean that my taste in music is "better" than anyone else's. 
(We've all met stupid lawyers.) But it does say on the one hand that I 
can probably tell you whether a given melody was lifted from Bach, even 
if it was transposed and inverted and the instrumentation was changed.
When I was in university I took a course called "History of Music." 
Doctor Ursula Rempel told us in the first class that the exam would be 
to listen to 20 fragments of music (each 10 seconds long), and we'd have 
to identify the type of work, the movement if possible, the composer if 
possible, and the year in which it was written (within 20 years). When 
the good doctor said that, I thought there's no way in the world I could 
possibly do that. This incidentally was a summer course; I attended 
class every day for 6 weeks. By exam time I thought that 10 seconds was 
an absurdly long time for each question. Ms. Rempel had taught me how to 
recognize Renaissance, Baroque, Rococo, Classical, Romantic, Late 
Romantic, Early Modern, 12-tone etc. almost instantly... and it was 
almost as easy to say this is German as opposed to Italian. A few trick 
questions could undermine you... for example, Schubert is pretty close 
to Beethoven, and Carl Czerny is even closer, and certain composers make 
a point of trying to confuse you with era. But the fact is that most of 
the time I can tell you (within the classical European, jazz and 
East-Indian classical music traditions) who is playing and what 
composition type is being played within seconds. If it's tricky, it 
might take me a minute, and if it takes me longer than that then I'm 
just guessing.
All the foregoing was about music from the dare I say it, educated 
listener's point of view. This axis has virtually nothing to do with 
what will sell. I like to think that I have an ear for quality (don't we 
all), and I have a certain amount of evidence to cite. Not a lot of said 
evidence concerns record sales, but rather longevity. There are things 
you can do in the world of European classical music that are impossible 
in other genres. For example, I have approximately 20 recordings of Igor 
Stravinsky's "Rite of Spring" and I can tell you without hesitation 
which one I consider the best, and even play fragments from the various 
versions to justify my position. You can't do that with rock, or movies, 
or most other musical genres. To a lesser extend you can do it with jazz.
So where am I going with this? I don't want to go into the corner thats 
says great music is only for those who know. Neither do I want to go 
into the corner that says that someone who has listened only to punk or 
rap or disco or classical Indian music can pronounce upon what is great 
music. I think that a LOT more perspective is required, and a much 
larger time-frame.
You proposed a much simpler proposition that is much easier to test. 
Let's just hope you don't come up with the musical equivalent of "Famous 
Dogs of the Civil War." Even if it does sell a jillion copies this year.

LOL. I do tend to ramble on.
A.
There's a rule in S-F writing circles: introduce exactly 1 radical new 
concept and base your book upon that. There are numerous exceptions to 
this rule: to cite just three, William Gibson, Neal Stephenson and 
Philip Kerr. But in general, I think that the rule holds. It doesn't 
guarantee success, by any means, but it does describe many and perhaps 
the majority of successful S-F novels.
I vaguely recall a joke about this sort of analysis, too. Some book 
publisher decided to search for the three most successful themes in 
novels, thereby to derive the formula for the next blockbuster, and 
after all the data was sifted and the numbers crunched, the software 
proposed "Famous Dogs of The Civil War."
So in fact, your perspective (I think -- don't let me put words in your 
mouth), you nest two other questions and possibly three. IOW, you 
identify one axis as the measure of the database's success: future sales 
of the proposed artwork. That's fine, as far as it goes, but I think it 
does not go very far... except, assuming success, all the way to the 
bank. What I must applaud about this approach is its scientific 
perspective (i.e. prediction and control) -- you propose a case that can 
be tested objectively in a relatively small time-frame, whereas lofty 
frames of reference such as "greatness", "beauty", "influence over 
subsequent composers" etc. require much more subjectivity and much 
larger time-frames.



More information about the dba-Tech mailing list