337 -- Intro
to Semantic Information Processing-- L. Birnbaum
CAUSAL CHAINS AND INFERENCE, II
Last time, we concluded that the
representation of a coherent
paragraph describing some event consists of a
CAUSAL CHAIN that links
the individual actions and states involved.
It seems intuitively
clear that understanding involves building
such a representation,
since it is a direct outcome of, and reflects,
our ability to explain
Behaviorally, what can we use causal chains
To answer questions:
Who gave the slave girl the necklace?
To summarize stories:
Jack was thirsty. He went to
the kitchen and took a beer from the
refrigerator. His wife asked
him to do the dishes. He told her
he would do them when he finished the beer.
He went to the den.
He turned on the radio. He
heard that the weatherman predicted it
would rain tomorrow. He went
to a chair and sat down. The chair
fell over and Jack landed on the floor.
The beer spilled all over
the chair. When Jack's wife
saw the mess, she was angry.
Summarization rules (Schank 1975):
1. Deadend chains will be forgotten.
2. Sequential chains may be shortened.
3. Disconnected pieces will be connected or forgotten.
4. Pieces that have many causal connections are crucial.
Summary: Jack was thirsty. He
got a beer. He went into the den.
He sat down on a chair. The
chair fell over. Jack fell on the
floor. The beer
spilled on the chair. Jack's wife
Let's turn to Charniak's (1972) theory.
Unlike Rieger and Schank, who
propose both a theory of content and of
process, Charniak's theory is
basically a process theory -- it makes no
commitment to the kinds of
inferences that will be drawn.
Charniak doesn't make all inferences all the
time. Some inferences
are made right away, others are deferred.
BASE ROUTINES attached to concepts make immediate inferences, and
DEMONS represent expectations -- they draw inferences only
the mention of certain other concepts -- i.e., there is a notion
Example: The base routine for "piggy
bank" will activate the following
1 If you want money, then you can get it from the piggy bank if
you get the piggy bank.
2 If you want to get money out of the piggy bank, shake it.
3 If you shake a piggy-bank, and you don't hear anything, then
there is no money in it.
So consider the following story:
Janet wanted to get some money. She
found her piggy bank and
started to shake it. She
didn't hear anything.
Demon 1 links sentence 1 to sentence 2, demon 2 operates within
sentence 2, demon 3 links sentence 2 to sentence 3.
assertions --> active demons --> base routines
Compare with Rieger:
1 No specific inference types -- i.e., no content theory of causal
2 Deferral of inference.
3 Demons are more goal-directed.
4 Handles inferences from complex conjunctions of features much
better -- i.e., the inferences are more SPECIFIC.
Now let's consider the following story:
John went to a restaurant.
He asked for a hamburger.
He paid and left.
How could we do this with Rieger inference?
From the first sentence, we can infer that was at the restaurant
(resultative), from which we can infer that John wanted to be at
the restaurant (motivational), from which we can infer that John
wanted to eat something (functional).
Along with many other,
mostly irrelevant inferences, e.g., John was somewhere else before
he went to the restaurant (enablement).
It's hard to know how much we can expect to get out of the second
sentence "bottom-up", looking only at the words it contains.
Giving Rieger the benefit of the doubt, we'll say it means "John
told someone that he wanted someone to give him a hamburger." From
this we can infer that John wanted someone to give him a hamburger
(enablement), from which we can infer (maybe) that John wanted to
have a hamburger (what inference type would this be?), from which
we can infer that John wanted to eat a hamburger (functional).
This matches the inference from sentence 1 that John wanted to eat
something. So we know that
he asked for the hamburger for the
same reason that he went to the restaurant -- in order to eat.
This hardly captures what we know here -- which is that you ALWAYS
ask for what you want to eat at a restaurant.
"He paid" means that he gave someone some money in exchange for
that someone giving something to him.
Through some kludging, we
might be able to get this to match the fact that John wanted
someone to give him a hamburger.
Thus, with a little juggling, a program using
could be made to realize that John went to the
restaurant because he
wanted to eat, that he asked for the hamburger
for the same reason,
and, with a little more juggling, maybe even
that he ate the
hamburger, and that he paid for the hamburger.
This is rather unsatisfactory for a number of
PROCESS: Too many irrelevant inferences.
It would also infer such true but irrelevant gems as the fact
that John was somewhere else before he went to the restaurant,
that he had gone to that somewhere else, that someone gave him
the money that he used to pay, that someone gave that person
the money, etc.
Furthermore, of course, we can write restaurant stories that
leave even more implicit:
John went to a
The salad bar
He left 5
dollars on the table.
CONTENT: We should be using more specific knowledge.
It leaves many questions unanswered: DID John eat the
hamburger or not? It also
probably sat down.
That he may
have looked at a menu.
That a waiter
took his order.
That the waiter
told the cook his order.
That the waiter
brought the hamburger.
That the waiter
brought the check.
We also know that people ask for the food they want in
restaurants -- this is not an accidental relation.
the Rieger program would fail to recognize that the fact that
John went to the restaurant to eat, and that he asked someone
to bring him a hamburger so that he could eat, are JOINTLY
responsible for him getting something to eat.
That is, they
are RELATED elements in a single plan.
As far as Rieger's
program would know, the above restaurant story is no different
from the following:
John went to
He asked for a
He paid and