[PLUG-TALK] Linguistics anyone?

John Jason Jordan johnxj at gmx.com
Wed May 27 03:54:16 UTC 2020


Tomas,

What I have to say does not directly address the issues you mention.
However, you cannot even begin to work on your issues unless you know
a lot of the material that I do discuss. In other words, my information
is essential background. However, I never mention AI, nor do I discuss
programming computers to understand human language.

You mention one matter that I have long noted - human languages are
universally redundant. Consider the following sentence:

	Der Junge trägt seinen Hut.

Consider the fact that human languages are universally created to be
understood under far less than ideal conditions. If I said that
sentence to you out loud standing right in front of you your ears
would hear only about 40% of the sounds that I utter. Yet you would
understand me without a problem. How can that be?

We linguists consider our ability to understand is based on bottom-up
and top-down processing. Hearing the sounds that I utter is bottom-up
processing. But you know the language that I wrote that in, and if you
heard me say it you would guess most of the sounds that you did not
hear. Furthermore, from your knowledge of the language you might guess
the last word, Hut, from the ending on seinen, and if you only heard
'Ju' from Junge you could probably fill in the rest of the word. 

Our brains and the rest of our language apparatus evolved to deal with
the fact that sometimes our lives depend on being able to understand an
utterance that was shouted at us from too far away. You want a computer
to be able to emulate that human ability to understand. Then your
computer has to be able to understand everything about the syntax of
the language that you are training it to understand. Add the phonology
of the language (also massively complex), and the morphology, and I
could go on and give you a list of the topics that I cover. 

I am guessing that you want to start at the computer programming end of
your task, but you can't start there. You have to start by
understanding how human languages work. I can teach you a lot about
that part of the job; but I can't help you with the programming part.

I should also add that several years ago at PSU I ran into a young man
(early 20s) who was a dual major - computer science and linguistics.
His goal was to work on exactly the kind of thing we are discussing.
When he graduated his starting salary was six figures, and he had lots
of corporations line up wanting to hire him. You are right to be
interested in this. It is a huge field today.

But you still need to understand the rules of syntax, phonology, and so
on before you can program the computer.



On Tue, 26 May 2020 20:01:58 -0700
Tomas Kuchta <tomas.kuchta.lists at gmail.com> dijo:

>The only thing I know about linguistic is that I do not know anything
>about it.
>
>That being said, being able to rationalize what parts of language carry
>information, what kind of information and what is redundant filler,
>possibly used for error correction - is currently extremely active
>area of computer science.
>
>Reasonably modern algorithms for text comprehension throw out the
>filler stuff, convert text or speech to vectors and attempt to
>understand it.
>
>If what you do has something to say about that, it might be interesting
>topic for people interested in CS.
>
>Language design is also very interesting and still active area of CS.
>Sadly, not many languages are designed to not be ambiguous and  highly
>expressive.
>
>Just a thought,




More information about the PLUG-talk mailing list