| View previous topic :: View next topic |
| Author |
Message |
peter
Guest
|
 Posted: Sat Nov 15, 2008 9:07 am Post subject: nlp project |
 |
|
A guy are paying me to develop a nature language processing eninge, to
convert english into logic. I think I can write a phraser to phrase
the english. My idea is to convert english into 3 basic logics :
equal, larger than, small than.
If input "cat is red" into my engine, I will store "cat = red" in my
database.
So what should I do next?
thanks
from Peter (cmk128@hotmail.com)
[ comp.ai is moderated ... your article may take a while to appear. ] |
|
| |
|
 Back to top |
Carsten van Weelden
Guest
|
| Posted: Sun Nov 16, 2008 7:59 am Post subject: Re: nlp project |
|
|
On Nov 15, 5:45 am, peter <cmk...@hotmail.com> wrote:
| Quote: | A guy are paying me to develop a nature language processing eninge, to
convert english into logic. I think I can write a phraser to phrase
the english. My idea is to convert english into 3 basic logics :
equal, larger than, small than.
If input "cat is red" into my engine, I will store "cat = red" in my
database.
So what should I do next?
thanks
from Peter (cmk...@hotmail.com)
|
The more usual approach is to use a formal meaning representation,
first order predicate calculus is a good example based on first order
logic and lambda calculus. For instance, if you have the phrase "a red
cat" you would get something like Ex Cat(x) & HasColor(x, red).
You should start by finding out exactly what you have to represent. Do
you need to represent time or belief, then you need temporal or modal
logic. Will all the sentences concern a certain domain? Then you can
use a domain ontology to specify useful categories. Will some events
be very common? For example, if all the sentences are concerned with
making a reservation, then you know which aspect to represent and how
to store them in a database.
[ comp.ai is moderated ... your article may take a while to appear. ] |
|
| |
|
| Back to top |
HaltingState
Guest
|
| Posted: Sun Nov 16, 2008 8:01 am Post subject: Re: nlp project |
|
|
peter wrote:
| Quote: | A guy are paying me to develop a nature language processing eninge, to
convert english into logic. I think I can write a phraser to phrase
the english. My idea is to convert english into 3 basic logics :
equal, larger than, small than.
If input "cat is red" into my engine, I will store "cat = red" in my
database.
So what should I do next?
thanks
from Peter (cmk128@hotmail.com)
|
Notice that a relationship between objects is undirected if f(x,y) =
f(y,x); but for most relationships this is not the case.
[ comp.ai is moderated ... your article may take a while to appear. ] |
|
| |
|
| Back to top |
David Kinny
Guest
|
| Posted: Sun Nov 16, 2008 8:56 am Post subject: Re: nlp project |
|
|
HaltingState <HaltingState@gmail.com> writes:
| Quote: | peter wrote:
A guy are paying me to develop a nature language processing eninge, to
convert english into logic. I think I can write a phraser to phrase
the english. My idea is to convert english into 3 basic logics :
equal, larger than, small than.
|
These are relationships, not logics. You'll need more than these 3
to capture even a small fragment of English.
| Quote: |
If input "cat is red" into my engine, I will store "cat = red" in my
database.
So what should I do next?
|
Refund the guy's money?
| Quote: |
thanks
from Peter (cmk128@hotmail.com)
Notice that a relationship between objects is undirected if f(x,y) =
f(y,x); but for most relationships this is not the case.
|
The usual terminology is that such a relationship is symmetric.
(Equivalence relations are reflexive, symmetric and transitive.)
Also, red is not an object but a property, and "is" doesn't usually
stand for equivalence, e.g. a dog is a mammal, today is hot, today
is Sunday, Peter is smart, Peter is 20, Peter is early, ...
David
[ comp.ai is moderated ... your article may take a while to appear. ] |
|
| |
|
| Back to top |
Wolfgang Lorenz
Guest
|
| Posted: Sun Nov 16, 2008 5:36 pm Post subject: Re: nlp project |
|
|
peter wrote:
| Quote: | If input "cat is red" into my engine, I will store "cat = red" in my
database.
So what should I do next?
|
1) Input "rose is red" into your engine.
2) To verify that it works correctly, ask the database "cat is red?",
"rose is red?", and "blood is red?". It should answer "Yes", "Yes", and
"I don't know".
3) Ask the database "cat is rose?" What does it answer?
--
http://home.arcor.de/w.lorenz65/mlbench
No mercy for the cheaters in machine learning!
[ comp.ai is moderated ... your article may take a while to appear. ] |
|
| |
|
| Back to top |
peter
Guest
|
| Posted: Mon Nov 17, 2008 9:07 am Post subject: Re: nlp project |
|
|
Wolfgang Lorenz <wl924...@arcor.de> wrote:
| Quote: | peter wrote:
If input "cat is red" into my engine, I will store "cat = red" in my
database.
So what should I do next?
1) Input "rose is red" into your engine.
2) To verify that it works correctly, ask the database "cat is red?",
"rose is red?", and "blood is red?". It should answer "Yes", "Yes", and
"I don't know".
3) Ask the database "cat is rose?" What does it answer?
--http://home.arcor.de/w.lorenz65/mlbench
No mercy for the cheaters in machine learning!
|
Thank you guys
I think I am better to do it from the ground up. I am writing a
phraser using gcc with bison, the first goal is the define all the
english grammar rules. After that, I will input some books into the
engine, to see what logic it can collect.
thanks
from Peter
[ comp.ai is moderated ... your article may take a while to appear. ] |
|
| |
|
| Back to top |
|
