jpolowin at hotmail.com
Mon Dec 13 20:23:00 GMT 2021
Raymond Collins <rcrcoll6 at gmail.com> wrote:
> Dementia is a terrible thing. I had friends who've I've lost to dementia. I
> just hope things won't get worse. All that is great about a person seems to
> diminish with dementia, and it's only the reminiscence of who that person
> was is what we should hold on to the most. I hope, under the
> circumstances, that your Mother-in-law finds peace.
A few years ago, I went to a memorial service for one of my high-school
teachers. Her family told us that she had been suffering for a couple
of decades from Alzheimer's, which had been expressed in a particularly
ugly form. It was only after a couple of months that her family had
been able to really start remembering the wonderful mother, artist,
There is a thing that I want to exist. I don't have the skills to do
it, nor the spoons to try to organize it.
My sweetie has been hospitalized many times over the decade that we've
been together. A number of times, she's been stuck sharing a room
with someone who's been confused from dementia. Even when she hasn't,
it's almost always possible to hear someone yelling in some nearby
room, and clearly suffering from some kind of dementia. They don't
understand where they are, much less why they're there. A nurse can try
to re-orient them, but that only lasts a few minutes, since they can't
retain the information.
One time, my sweetie was sharing her room with an extremely frail old
woman, who kept calling out to "Margaret" in her weak whispery voice.
"Is my oxygen on? Margaret? Are you there? Is my oxygen okay?" It
was heartbreaking. (Notwithstanding that my twisted brain immediately
jumped to "Are you there, Margaret? It's me, God.") Our attempts to
reassure her lasted only minutes.
It occurred to me that what she *really* needed was some kind of
automated system that would recognize the spoken word "oxygen" and tell
her that it was fine.
And from there came the idea. A voice recognition, or speech
recognition, that could parse what such a patient was saying, just
enough to pick out a few words or phrases, and reply with reassurance.
"Where am I?" or "What's going on?" could respond with "You're in
the hospital and you're going to be okay." Do it with the voice of a
loved one, recorded for the purpose. Parsing the voice of a patient
would be complicated by the difficulty of getting consistent sound, the
impossibility of doing "training" of the system with a standard set of
phrases, and poor enunciation on the part of the patient. The problem
would be simplified by not needing to recognize a large vocabulary;
it wouldn't necessarily even need to pick out *words* as long as sound
patterns were consistent enough -- if a mumbled "oxygen" always sounds
about the same, it would be enough to recognize the sound pattern.
It shouldn't require fancy hardware, by current standards. This is
basically an audible version of the old "Eliza" programs: scan the
input for patterns, and choose a stock sound clip to play in response.
There's no great AI leap needed. It seems to me that it should be
possible to run it on a fairly cheap phone, tablet, or laptop.
It would save a lot of effort on the part of nurses, and others who have
to care for people with severe dementia. It would also provide a great
deal of comfort to the patients. I've run the idea past nurses, and
they love it. My sweetie, a dementia specialist herself, thinks it's
a great idea. She thinks it might be possible to get funding for the
development from Alzheimer's and other dementia foundations. I'd like
to see the thing available for free, at least as far as the software
goes, and working on as many different platforms as possible. Make it
as available as possible for people who don't have a lot of resources.
If anyone can suggest what I can do towards getting someone to do
something with the concept, I'd appreciate it. I just have almost no
"push" left in me. Basically, I'd like to hand the idea off to someone
who will recognize it as being in the public interest, and get it done.
More information about the Lois-Bujold