Title: On database keys, with an application to the Praxisproblem Derek J' SMITH Centre for Psychology Univ
1On database keys, with an application to the
PraxisproblemDerek J. SMITHCentre for
PsychologyUniversity of Wales Institute,
Cardiffdsmith_at_uwic.ac.ukhttp//www.smithsrisca.
demon.co.uk
2As presented to the9th World Multi-Conferenceon
Systemics, Cybernetics, and InformaticsOrlando,
FL, Tuesday 12th July 2005
3A BIT MORE ABOUT THE AUTHOR
- 1980s - specialized in the design and operation
of very large IDMS databases. - Since 1991 taught cognitive science and
neuropsychology to Speech and Language
Pathologists. - Hence interdisciplinary in database, cognitive
neuropsychology, and psycholinguistics. - Hence fascinated by the possibility that the mind
is a biological database capable of linguistic
communication.
4DEFINITIONS (1)VOLITION AND PRAXIS
- Volition is the exercise of the Will.
- Praxis comes from the Greek prassein, "to do,
and refers generally to a systems ability to
initiate voluntary physical behavior of any sort
(hence we call it dyspraxia when something goes
wrong). - Speech praxis is the process by which the Will
voluntarily initiates the expression of a given
idea though the medium of spoken language.
5DEFINITIONS (2)REDUCTIONISM
- Reductionism is explaining complex phenomena in
terms of their parts. E.g. how nuts, bolts, etc.,
make an automobile. - However, theres a big problem with reductionist
explanations, because systems are frequently
greater than the sums of their parts, and you
lose much of their essence as soon as you start
taking them apart. This is especially true of the
mind, which seems to be massively more than the
sum of its neurons. - Following Levine (1983), we favor the term
Explanatory Gap for this problem.
6PLAN OF ATTACK
- This paper looks at the explanatory gap problem
in speech praxis, and shows how that gap might be
narrowed in an artificial cognitive system using
CODASYL-style database keys. - This means bringing together two hitherto
unconnected areas of knowledge, so we shall open
with background" sections on (a) the
psycholinguistics of speech praxis, and (b) the
use of database keys in CODASYL databases. Well
then present a pseudocode demonstration of how
the one can help the other.
7Well also be telling you a little about what
went on at this place a while back.
8LORDAT (1843)FIVE STAGES OF SPEECH
PRAXIShttp//www.smithsrisca.demon.co.uk/PSYlorda
t1843.html
- So what do psycholinguists know about speech
praxis? - Well in 1843 the French neurologist Jacques
Lordat identified five processing stages within
speech production, and described the first of
these as isolating the idea to be expressed in
words. - The four subsequent processes then co-operate in
shaping the final spoken output, as now shown ..
9LORDAT (1843)
- Each stage receives a coded message from the one
before, adds to it in some clever way, and then
passes it on to the one after.
10LORDAT (1843)FIVE STAGES OF SPEECH PRAXIS
- Lordats explanation was duly incorporated into a
number of later 19th century aphasiological
models, and is worth noting because in one form
or another it is still with us today. - The explanatory diagrams are typically either
A-shaped (with the clever bits the minds
higher functions - at the top) or X-shaped
(with the clever bits at the central cross-over
point). Here are two of the A-shaped ones, one
old and one new ..
11LICHTHEIM (1885)TWO-LAYER CONTROL
HIERARCHYhttp//www.smithsrisca.demon.co.uk/PSYli
chtheim1885.html
12NORMAN (1990)MODERN THREE-LAYER
HIERARCHYhttp//www.smithsrisca.demon.co.uk/PSYno
rman1990.html
13MODERN SPEECH PRODUCTION MODELS
- Speech production did not become a popular study
topic until 1971, when Fromkin's "Utterance
Generator" model extended Lordat from five stages
to six. - The most popular modern models of speech
production come from the Max Planck Institute's
Willem Levelt and the University of Arizona's
Merrill F. Garrett. But they all take the same
basic shape, so well work for now with Fromkins
..
14FROMKINS FIRST THREE STAGES
- Stage 1 Pre-Lexical Semantics Decides the
meaning to be conveyed. Code not known, but
preverbal. - Stage 2 Pre-Lexical Syntax Decides the
grammatical skeleton of the sentence. Code not
known, but preverbal. - Stage 3 Lexical Selects the necessary content
words (i.e. nouns and verbs) from the mental
lexicon, thus making ideas verbal for the first
time. - Its worth remembering these three stages as a
unit, because well be returning to them later.
15FROMKINS LAST THREE STAGES
- Stage 4 Prosody Adds in emotionality via
intonation pattern. Code not known, but mediated
by the hindbrain. - Stage 5 Phonology Decides the final syntax and
word morphology. Phonemic code. - Stage 6 - Final Sound Production Commits
concrete sounds - "allophones to the motor
nerves for respiration, phonation, and
articulation. - Were not really concerned with these three
stages in this paper, because they are all
post-semantic.
16BUT ARE PROCESSING STAGES THE SAME AS MODULAR
LAYERS?
- Theres another big problem here, because
Fromkin's six stages do not necessarily map onto
six CPUs each running a single computer program.
The diagrams we saw just now are all showing
processors physical modules - not processing
stages. - In fact, its a many-to-many relationship. You
can fit Fromkin's six stages perfectly into
Lichtheim's two physical processing modules if
only you bundle them properly! This many-to-many
cross-mapping has never been properly resolved,
and is still a major source of mis-understanding
and confusion within Artificial Intelligence.
17STATE-OF-THE-ARTPSYCHOLINGUISTIC MODELING
- The PALPA (Kay, Lesser, and Coltheart, 1992) is a
typical X-shaped psycholinguistic diagram (that
is to say, the Will is located in the middle of
the diagram, not at the top). - Here it is ..
18The PALPAhttp//www.smithsrisca.demon.co.uk/PSYka
yetal1992.html
- All input channels are now at the top, and all
output channels at the bottom. Ideation is in the
centre box, and speech praxis is the bottom left
processing leg. - Our question is, how do ideas make it out of the
centre box and down the first arrow!
19SPEECH ACTS AND IDEATION (1)
- In fact, praxis is so complex that it has its own
science pragmatics .. - .. and its own very powerful theory - Speech Act
Theory (Austin, 1962 Searle, 1969). Speech Act
Theory studies not just the words people use, but
the units of intention the speech acts which
preceded those words. - Pragmatics is thus the science of the centre box
and the first arrow ..
20SPEECH ACTS AND IDEATION (2)
- Each speech act is (a) calculated to achieve some
discrete behavioral "perlocutionary" effect, but
(b) has not yet been fully formed lexically or
grammatically. - The code is preverbal - perhaps sprites or
ideograms of some sort. - Speech acts havent been very well modeled yet,
and so remain poorly understood (although we hope
to start the ball rolling with this paper).
21SPEECH ACTS AND IDEATION (3)
- Speech Act Theory has its own very difficult
jargon. In this paper, for example, we shall be
hearing about a directive speech act, with a
requestive subtype, invoking the core verb to
summons. - The mind has about a 1000 different speech acts
to choose from (more details in Bach and Harnish,
1979).
22ANIMATED PALPA SMITH (2000)
- So what might a speech act look like? Where do
these all-important sprites come from, where do
they go, and what happens to them when they get
there? - To get a better idea of the process, we need to
see the static flow diagram in motion. So here,
from Smith (2000), is sentence production more or
less in real time, for the specimen sentence The
Redcoats are coming ..
23(No Transcript)
24(No Transcript)
25OUR BASIC PROPOSAL
- So how might a network database help?
- Well above all it would help demonstrate how
mental content is progressively accessed during
sentence production. Not just ideation and
initiation (Stages 1 and 2), but lexical
expansion (Stage 3) as well. - Heres some important background (weve taken
Computer Associates' IDMS system, a 1969 network
database, as the class-defining CODASYL database)
..
26ABOUT NETWORK DATABASES (1)IDMS DATABASE KEYS
- Data is organized into "records.
- Filestore is organized into block-sized "pages
of up to 256 records per page. - The resulting page-line address is known as the
"database key" for that record.
27ABOUT NETWORK DATABASES (2)IDMS DIRECT ACCESS
- IDMS is a supremely flexible data retrieval tool.
It offers 10 ways of finding things, the most
fundamental of which is direct access. Just like
the brain during every act of perception.
28ABOUT NETWORK DATABASES (3)IDMS DIRECT ACCESS
- A hashing algorithm computes the page number from
the logical record key (Dumey, 1956). - Once a record has been located, its database key
is automatically stored by the DBMS as a
currency.
29ABOUT NETWORK DATABASES (4)CHAIN POINTER ACCESS
- IDMS also arranges for certain data records to
"own" certain others, thus capitalizing on the
data's natural one-to-many "set" relationships.
30ABOUT NETWORK DATABASES (5)CHAIN POINTER ACCESS
- Each record includes in its physical record
length the database key of the next record in
that set. These chain pointers are a storage
overhead, but one which is well worth paying.
Just like the brain during semantic or lexical
categorization.
31ABOUT NETWORK DATABASES (6)CURRENT OF SET
ACCESS
- Once a record has been located, its database key
is automatically stored by the DBMS, identifying
it as current of set for the set in question. - This makes said record effectively direct
access, even if it had originally taken a lot of
chain pointering to track down. Just like the
brain during medium term recall.
32ABOUT NETWORK DATABASES (7)IDMS TRAVERSALS
- In practice, the final outputs of a database
interrogation are always composites of fragments
of data collected from potentially hundreds of
points across the network. The process of
collecting these fragments is known as
traversing the database. The inventor of the
process, Charles W. Bachman, winner of the 1973
A.C.M. Turing Award, once likened it to
"navigating" (Bachman, 1973). - Traversals are hard to plan, but then easy to
achieve using only a handful of "Data
Manipulation Language" (DML) verbs, such as
OBTAIN CALC, OBTAIN NEXT, MODIFY, and ERASE.
33TOWARDS AN IDMS "MINDBASE" (1)DATABASE KEYS AND
MACHINE VOLITION
- OK, so lets start to bring these thoughts
together .. - Supposing the mind were an IDMS database, then
for database traversal purposes, only the first
three of Fromkin's six stages would need to be
considered. This is because they are the ones
which access the semantic network. Everything
else is just number-crunching. - Lets look at those stages one by one ..
34TOWARDS AN IDMS "MINDBASE" (2)ISOLATING A
PROPOSITION
- Stage 1 is where we tap into the minds central
stream of contextually coherent propositional
thought. - Each successive proposition will activate a
number of the mental sprites we have been talking
about. Always ltagentgt and ltactiongt, and sometimes
also ltobjectgt, ltadverbialgt, and ltinstrumentalgt. - Every now and then, a small subset of
propositions becomes salient enough to warrant
initiating a communicative act, whereupon that
particular unit of ideation is "isolated" ready
for production.
35TOWARDS AN IDMS "MINDBASE" (3)SPEECH ACT
SELECTION
- Stage 2 is responsible for converting the
afore-mentioned subset of propositions into a
speech act of some sort. - The point is that by selecting a particular
speech act, we restrict our choice of possible
sentence structures to only a handful of options.
It is rather like opening sealed orders on a
military mission - one specific action gets
selected from a repertoire of many, you get told
what to do next, and the desired outcomes are
specified.
36TOWARDS AN IDMS "MINDBASE" (4)LEXICAL EXPANSION
- Stage 3 is where we start attaching words to
what until now have been preverbal codes. - Unfortunately, this process does not always work
cleanly see our e-paper on Speech Errors at
www.smithsrisca. demon.co.uk/speech-errors.html.
37TOWARDS AN IDMS "MINDBASE" (5)STAGES 1 TO 3 IN
PSEUDOCODE
- Here are some extracts from the pseudocode set
out in the written version of this presentation.
The selections show the following four
circumstances .. - (1) Isolation of ideation.
- (2) Selection of a speech act by direct access.
- (3) Chain pointer use in opening a set.
- (4) Set currency use in continuing a set.
38INDICATIVE IDMS PSEUDOCODE (1)ISOLATING PAUL
REVERES IDEA
- Different combinations of external circumstances
demand different actions .. - IF AGENT .
- .
- ELSE IF AGENT BRITISH COLUMN
- IF ACTION On null condition ..
- NEXT-SENTENCE .. no action is necessary
- ELSE IF ACTION MOVING NORTHWEST
- MOVE PLAN X TO SPEECH-ACT-CODE
- NEXT SENTENCE
- ELSE IF .
- ELSE IF .
- .....
39INDICATIVE IDMS PSEUDOCODE (2) EXAMPLE OF DIRECT
ACCESS
- Once a given set of circumstances has prompted a
particular pre-learned plan of action, the
detailed response can be retrieved from long-term
memory .. - MOVE SPEECH-ACT-CODE TO ACT-RECORD-KEY.
- OBTAIN CALC ACT-CONTROL-RECORD.
- The sealed orders on this occasion specify a
directive speech act, with a requestive
subtype, and the core verb to summons.
40INDICATIVE IDMS PSEUDOCODE (3) EXAMPLE OF CHAIN
POINTER ACCESS
- The waypoints on the road to Lexington are
themselves in long-term memory, organized on an
owner/member set basis .. - MOVE MIDDLESEX COUNTY TO DISTRICT-KEY.
- OBTAIN CALC DISTRICT. We dont actually need this
record, but we have to get it because it contains
the chain pointer to the first set member .. - OBTAIN FIRST MINUTEMAN IN MINUTEMEN-IN-DISTRICT.
41INDICATIVE IDMS PSEUDOCODE (4) EXAMPLE OF SET
CURRENCY USE
- Having processed one minuteman, we have to
relocate its record before getting the next,
again because we need the chain pointer it
contains .. - OBTAIN CURRENT MINUTEMAN.
- OBTAIN NEXT MINUTEMAN IN MINUTEMEN-IN-DISTRICT.
42CONCLUSION (1)WHAT WE HAVE SEEN SO FAR
- This has been no more than a first foray into the
explanatory gap, in an attempt to redefine the
philosophical problems of biological cognition as
network database problems. - The pseudocode is entirely typical of IDMS
programming, in that you have to pay precise
attention (a) to where you are in the data
network, and (b) to how you are going to get to
where you want to be next. The record and set
currency mechanisms are invaluable in this
respect. - We may identify three specific potential
advantages of the IDMS metaphor ..
43CONCLUSION (2)IDMS IS LIKE FOR LIKE WITH BIOLOGY
- The network database is just that, a network, and
so, too, is biological associative memory. - And yet only a microscopic proportion of the
billions spent on AI in the past 50 years has
utilized a physical network architecture.
44CONCLUSION (3)IDMS IS TOTALLY REDUCIBLE
- The activity of an IDMS mindbase is automatically
totally reducible. The DML is pre-processed to
native COBOL, and then put through the regular
COBOL compiler to produce object code. Both
conversions are 100 trackable, and in debug mode
we can even execute the object code one
instruction at a time. - This is important because what cognitive
theorists are really about is reverse engineering
- de-compiling neural object code in order to
understand what on earth it is up to.
45CONCLUSION (4)IDMS MIMICS NEUROCHEMISTRY
- IDMS's use of database currencies is uncannily
similar to the brain's physiological memory
mechanisms. E.g. second messenger
neurotransmission and calcium switching. In
other words, biological long-term memory looks to
have its own keys and currencies (Smith, 1997). - Or to put it another way, closing the explanatory
gap is a code-breaking exercise of sorts!
46CONCLUSION (5)CRYPTANALYSIS IN COGNITIVE SCIENCE
- The building we saw earlier was Station X
Bletchley Park The UK Code and Cipher School
during World War Two, where Alan Turing helped
decode the Nazi Enigma signal system ..
47CONCLUSION (6)THE COLOSSUS
- .. and the birthplace in December 1943 of the
worlds first programmable electronic binary
digital computer.
48CONCLUSION (7)THE COLOSSUS BUILDING
- .. which was successfully rebuilt by
preservation enthusiasts between 1994 and 1996.
49CONCLUSION (8)TURINGS OFFICE (HUT 8) NOWADAYS
- The same Alan Turing, incidentally, who inspired
the A.C.M. Turing Award, saw the mind as
computable, and devised the Turing Test of
machine consciousness.
50CONCLUSION (9)SPECIFIC PROPOSAL
- The best way across the explanatory gap would be
to approach it as a code-breaking exercise,
starting with neural crackle as the given
intercept, and reverse engineering it back to
the DML of the original ideation .. - .. and looking out especially for the location
of the biological currency tables. - IDMS may even assist students of artificial
consciousness at the same time ..
51CONCLUSION (10)ARTIFICIAL CONSCIOUSNESS
- A leading consciousness theorist, David Chalmers,
has argued that if a technically advanced theory
could one day explain at the reductionist level
how the brain thinks, it would still be "silent
about how these processes might give rise to
conscious experience" (Chalmers, 1995, p64). - We submit IDMS as a candidate for Chalmers'
"technically advanced theory .. - .. and suspect that its reducibility and
capacity for slow motion execution will reflect
on the how question as well.
52THE END
53REFERENCES
- Austin, J.L. (1962). How to do Things with Words.
Oxford Oxford University Press. - Bach, K. and Harnish, R.M. (1979). Linguistic
Communication and Speech Acts. Cambridge, MA
MIT Press. - Chalmers, D.J. (1995). The puzzle of conscious
experience. Scientific American, 273(6)62-68.
Levine (1983) - Fromkin, V.A. (1971). The non-anomalous nature of
anomalous utterances. Language, Vol. 47, pp.
27-52. - Searle, J.R. (1969). Speech Acts An Essay in the
Philosophy of Language. Cambridge Cambridge
University Press. - Smith, D.J. (1997). The IDMS Set Currency and
Biological Memory. Cardiff UWIC. ISBN
1900666057 - Smith, D.J. (2000). A slow-motion video analysis
of information feedback in a computer-animated
psycholinguistic model. Computer-animated poster
presented 10th April 2000 at the Tucson 2000 -
Towards a Science of Consciousness conference,
University of Arizona, Tucson, AZ. Available
online at http//www.uwic.ac.uk/shss/djs/palpa.avi