US speech \"accents\"...

[Lasse Langwadt Christensen]

fredag den 14. oktober 2022 kl. 00.40.09 UTC+2 skrev Dimiter Popoff:

On 10/14/2022 0:44, Lasse Langwadt Christensen wrote:
torsdag den 13. oktober 2022 kl. 23.12.03 UTC+2 skrev Dimiter Popoff:
On 10/13/2022 20:41, Lasse Langwadt Christensen wrote:
torsdag den 13. oktober 2022 kl. 19.24.29 UTC+2 skrev Jeff Layman:
On 12/10/2022 21:04, Phil Hobbs wrote:
Jeff Layman wrote:
On 12/10/2022 15:25, John Larkin wrote:
On Wed, 12 Oct 2022 01:46:18 -0700, Don Y
blocked...@foo.invalid> wrote:

Any comments (from US, english-native-speakers) re: regional
(english language) accents that are easiest and hardest to
understand? I.e., a single-pass being all you hear?

Texan, especially west texan and Oklahoma, can be tricky. And there is
hillbilly, Tennessee, which can get intense.

I am a Brit who went on holiday to New England about 30 years ago,
staying at a small B&B in Chester, Vermont. At breakfast one morning
three new guests appeared. We exchanged small talk. I understood very
little of what they said (no doubt they had trouble understanding me
too). After breakfast I asked the B&B owner where they were from, and he
said \"Oklahoma\". I said that I was somewhat embarrassed to admit I
couldn\'t understand what they were saying. He replied that he also
couldn\'t understand much of what they said!


Hell, in Britain, Glaswegian or Geordie is hardly less different from
Received Pronunciation.
I think that this thread is confusing accent and dialect. It is true
that Glaswegian can be very difficult to understand, but that is as much
due to them using their own words for certain things rather than just
having a very different pronunciation of words which are the same as
anywhere else in the UK.

I find Geordie the UK accent I like the most (and I\'m a Londoner by
birth), and other than some oddities don\'t usually have a great problem
understanding it.

I find it interesting is noticing Americans that complain they can\'t understand some
UK accents/dialects, and wondering why when I with English as my second language
find it perfectly understandable



Can you understand Glaswegian? I only get every other word at best when
I listen to the likes of Alex Ferguson or Kenny Dalglish, and they speak
of football, i.e. the words they use are not that many nor that fancy.

just checked a few interviews with them on youtube, doesn\'t seem to bad

Your filters must be really good and adaptive. I do understand parts of
what they say at times, especially when I know the context, but
typically I miss words and if they are key to the sentence I am just
lost.
Sometimes I get the feeling Glaswegians don\'t understand each other,
they just say their say and just assume what the others are saying .

like this old Italian song that\'s made to sound English but is just gibberish

https://youtu.be/-VsmF9m_Nt8

 

[Dimiter_Popoff]

On 10/14/2022 1:52, Lasse Langwadt Christensen wrote:

fredag den 14. oktober 2022 kl. 00.40.09 UTC+2 skrev Dimiter Popoff:
On 10/14/2022 0:44, Lasse Langwadt Christensen wrote:
torsdag den 13. oktober 2022 kl. 23.12.03 UTC+2 skrev Dimiter Popoff:
On 10/13/2022 20:41, Lasse Langwadt Christensen wrote:
torsdag den 13. oktober 2022 kl. 19.24.29 UTC+2 skrev Jeff Layman:
On 12/10/2022 21:04, Phil Hobbs wrote:
Jeff Layman wrote:
On 12/10/2022 15:25, John Larkin wrote:
On Wed, 12 Oct 2022 01:46:18 -0700, Don Y
blocked...@foo.invalid> wrote:

Any comments (from US, english-native-speakers) re: regional
(english language) accents that are easiest and hardest to
understand? I.e., a single-pass being all you hear?

Texan, especially west texan and Oklahoma, can be tricky. And there is
hillbilly, Tennessee, which can get intense.

I am a Brit who went on holiday to New England about 30 years ago,
staying at a small B&B in Chester, Vermont. At breakfast one morning
three new guests appeared. We exchanged small talk. I understood very
little of what they said (no doubt they had trouble understanding me
too). After breakfast I asked the B&B owner where they were from, and he
said \"Oklahoma\". I said that I was somewhat embarrassed to admit I
couldn\'t understand what they were saying. He replied that he also
couldn\'t understand much of what they said!


Hell, in Britain, Glaswegian or Geordie is hardly less different from
Received Pronunciation.
I think that this thread is confusing accent and dialect. It is true
that Glaswegian can be very difficult to understand, but that is as much
due to them using their own words for certain things rather than just
having a very different pronunciation of words which are the same as
anywhere else in the UK.

I find Geordie the UK accent I like the most (and I\'m a Londoner by
birth), and other than some oddities don\'t usually have a great problem
understanding it.

I find it interesting is noticing Americans that complain they can\'t understand some
UK accents/dialects, and wondering why when I with English as my second language
find it perfectly understandable



Can you understand Glaswegian? I only get every other word at best when
I listen to the likes of Alex Ferguson or Kenny Dalglish, and they speak
of football, i.e. the words they use are not that many nor that fancy.

just checked a few interviews with them on youtube, doesn\'t seem to bad

Your filters must be really good and adaptive. I do understand parts of
what they say at times, especially when I know the context, but
typically I miss words and if they are key to the sentence I am just
lost.
Sometimes I get the feeling Glaswegians don\'t understand each other,
they just say their say and just assume what the others are saying .

like this old Italian song that\'s made to sound English but is just gibberish

https://youtu.be/-VsmF9m_Nt8

Hah! Not dissimilar indeed . Strangely I had never heard that song,
it is not by just anyone, Celentano was quite a star.

 

[Lasse Langwadt Christensen]

fredag den 14. oktober 2022 kl. 00.59.28 UTC+2 skrev Dimiter Popoff:

On 10/14/2022 1:52, Lasse Langwadt Christensen wrote:
fredag den 14. oktober 2022 kl. 00.40.09 UTC+2 skrev Dimiter Popoff:
On 10/14/2022 0:44, Lasse Langwadt Christensen wrote:
torsdag den 13. oktober 2022 kl. 23.12.03 UTC+2 skrev Dimiter Popoff:
On 10/13/2022 20:41, Lasse Langwadt Christensen wrote:
torsdag den 13. oktober 2022 kl. 19.24.29 UTC+2 skrev Jeff Layman:
On 12/10/2022 21:04, Phil Hobbs wrote:
Jeff Layman wrote:
On 12/10/2022 15:25, John Larkin wrote:
On Wed, 12 Oct 2022 01:46:18 -0700, Don Y
blocked...@foo.invalid> wrote:

Any comments (from US, english-native-speakers) re: regional
(english language) accents that are easiest and hardest to
understand? I.e., a single-pass being all you hear?

Texan, especially west texan and Oklahoma, can be tricky. And there is
hillbilly, Tennessee, which can get intense.

I am a Brit who went on holiday to New England about 30 years ago,
staying at a small B&B in Chester, Vermont. At breakfast one morning
three new guests appeared. We exchanged small talk. I understood very
little of what they said (no doubt they had trouble understanding me
too). After breakfast I asked the B&B owner where they were from, and he
said \"Oklahoma\". I said that I was somewhat embarrassed to admit I
couldn\'t understand what they were saying. He replied that he also
couldn\'t understand much of what they said!


Hell, in Britain, Glaswegian or Geordie is hardly less different from
Received Pronunciation.
I think that this thread is confusing accent and dialect. It is true
that Glaswegian can be very difficult to understand, but that is as much
due to them using their own words for certain things rather than just
having a very different pronunciation of words which are the same as
anywhere else in the UK.

I find Geordie the UK accent I like the most (and I\'m a Londoner by
birth), and other than some oddities don\'t usually have a great problem
understanding it.

I find it interesting is noticing Americans that complain they can\'t understand some
UK accents/dialects, and wondering why when I with English as my second language
find it perfectly understandable



Can you understand Glaswegian? I only get every other word at best when
I listen to the likes of Alex Ferguson or Kenny Dalglish, and they speak
of football, i.e. the words they use are not that many nor that fancy.

just checked a few interviews with them on youtube, doesn\'t seem to bad

Your filters must be really good and adaptive. I do understand parts of
what they say at times, especially when I know the context, but
typically I miss words and if they are key to the sentence I am just
lost.
Sometimes I get the feeling Glaswegians don\'t understand each other,
they just say their say and just assume what the others are saying .

like this old Italian song that\'s made to sound English but is just gibberish

https://youtu.be/-VsmF9m_Nt8
Hah! Not dissimilar indeed . Strangely I had never heard that song,
it is not by just anyone, Celentano was quite a star.

afaiu it was quite a hit, but it was before I was born

 

[Anthony William Sloman]

On Friday, October 14, 2022 at 4:02:13 AM UTC+11, Fred Bloggs wrote:

On Wednesday, October 12, 2022 at 10:25:35 AM UTC-4, John Larkin wrote:
On Wed, 12 Oct 2022 01:46:18 -0700, Don Y <blocked...@foo.invalid> wrote:

<snip>

> You absolutely require subtitles to begin to understand people in Scotland, especially their ghetto people- this is aside from the fact they\'re not worth listening to.

I never did. And quite a few of them are well worth listening to. One of my colleagues from EMI Central Research (and eventually my boss there) acquired some 25 patents in the course of his career, and has given me more good advice than you could shake a stick at.

The Scots have always taken education more seriously than the English, and it showed up in the people I had contact with. Ian had got a Ph.D. at Edinburgh - on a the design of super-conducting electric motor big enough to drive an oil tanker. He was working on less massive electronics - for phase-array medical ultrasound - when we worked together

> Nearly the same for Australia, it\'s not the pronunciation so much as a bunch of words that just aren\'t used in the U.S.- slangy casual speak words.

This may depend on the size of your vocabulary. Mine is broader than most, and I did read a lot of books when I was young, some of them written by Americans and Scots.

--
Bill Sloman, Sydney

 

[Don Y]

On 10/13/2022 3:39 PM, Dimiter_Popoff wrote:

Your filters must be really good and adaptive. I do understand parts of
what they say at times, especially when I know the context, but
typically I miss words and if they are key to the sentence I am just
lost.

That\'s the problem with all \"accents\" (speech patterns foreign to YOUR
norm). One can learn to understand damn near anything. But, the effort
required to -- esp when it is an infrequent activity *or* where you
can\'t just ask for someone to repeat what they\'ve said -- can make
listening \"tedious\" or stressful.

Most synthetic speech suffers from this problem; it\'s OK in very small
doses *if* you\'ve become accustomed to it AND have a pretty good idea
of what is being said. (try listening to someone reading nonsense
syllogisms in a \"foreign accent\" and make note of your overall comprehension!)
But, it\'s not the sort of experience that you \"seek out\".

Users of The Reading Machine often complained of \"listener fatigue\",
despite the fact that they were THRILLED to have access to the materials
that were possible via the machine.

[I always used to say the voice could penetrate CONCRETE!]

Sometimes I get the feeling Glaswegians don\'t understand each other,
they just say their say and just assume what the others are saying .

 

[John Larkin]

On Thu, 13 Oct 2022 13:38:35 -0400, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

Jeff Layman wrote:
On 12/10/2022 21:04, Phil Hobbs wrote:
Jeff Layman wrote:
On 12/10/2022 15:25, John Larkin wrote:
On Wed, 12 Oct 2022 01:46:18 -0700, Don Y
blockedofcourse@foo.invalid> wrote:

Any comments (from US, english-native-speakers) re: regional
(english language) accents that are easiest and hardest to
understand?  I.e., a single-pass being all you hear?

Texan, especially west texan and Oklahoma, can be tricky. And there is
hillbilly, Tennessee, which can get intense.

I am a Brit who went on holiday to New England about 30 years ago,
staying at a small B&B in Chester, Vermont. At breakfast one morning
three new guests appeared. We exchanged small talk. I understood very
little of what they said (no doubt they had trouble understanding me
too). After breakfast I asked the B&B owner where they were from, and he
said \"Oklahoma\". I said that I was somewhat embarrassed to admit I
couldn\'t understand what they were saying. He replied that he also
couldn\'t understand much of what they said!


Hell, in Britain, Glaswegian or Geordie is hardly less different from
Received Pronunciation.

I think that this thread is confusing accent and dialect. It is true
that Glaswegian can be very difficult to understand, but that is as much
due to them using their own words for certain things rather than just
having a very different pronunciation of words which are the same as
anywhere else in the UK.

Okay, Aberdeen then.

I find Geordie the UK accent I like the most (and I\'m a Londoner by
birth), and other than some oddities don\'t usually have a great problem
understanding it.

Sure, but I can understand Okies and Cajuns just fine too. (And even
Canucks!)

Are you sure abooot that?

 

[Dimiter_Popoff]

On 10/14/2022 3:22, Don Y wrote:

On 10/13/2022 3:39 PM, Dimiter_Popoff wrote:

Your filters must be really good and adaptive. I do understand parts of
what they say at times, especially when I know the context, but
typically I miss words and if they are key to the sentence I am just
lost.

That\'s the problem with all \"accents\" (speech patterns foreign to YOUR
norm).  One can learn to understand damn near anything.  But, the effort
required to -- esp when it is an infrequent activity *or* where you
can\'t just ask for someone to repeat what they\'ve said -- can make
listening \"tedious\" or stressful.

Most synthetic speech suffers from this problem; it\'s OK in very small
doses *if* you\'ve become accustomed to it AND have a pretty good idea
of what is being said.  (try listening to someone reading nonsense
syllogisms in a \"foreign accent\" and make note of your overall
comprehension!)
But, it\'s not the sort of experience that you \"seek out\".

Users of The Reading Machine often complained of \"listener fatigue\",
despite the fact that they were THRILLED to have access to the materials
that were possible via the machine.

[I always used to say the voice could penetrate CONCRETE!]

Penetrate concrete with Don having trained the machine what to
say in test mode... well well.

 

[Don Y]

On 10/14/2022 3:54 AM, Dimiter_Popoff wrote:

On 10/14/2022 3:22, Don Y wrote:
On 10/13/2022 3:39 PM, Dimiter_Popoff wrote:

Your filters must be really good and adaptive. I do understand parts of
what they say at times, especially when I know the context, but
typically I miss words and if they are key to the sentence I am just
lost.

That\'s the problem with all \"accents\" (speech patterns foreign to YOUR
norm).  One can learn to understand damn near anything.  But, the effort
required to -- esp when it is an infrequent activity *or* where you
can\'t just ask for someone to repeat what they\'ve said -- can make
listening \"tedious\" or stressful.

Most synthetic speech suffers from this problem; it\'s OK in very small
doses *if* you\'ve become accustomed to it AND have a pretty good idea
of what is being said.  (try listening to someone reading nonsense
syllogisms in a \"foreign accent\" and make note of your overall comprehension!)
But, it\'s not the sort of experience that you \"seek out\".

Users of The Reading Machine often complained of \"listener fatigue\",
despite the fact that they were THRILLED to have access to the materials
that were possible via the machine.

[I always used to say the voice could penetrate CONCRETE!]

Penetrate concrete with Don having trained the machine what to
say in test mode... well well.

Shhhh! No one\'s supposed to know! :>

 

[Rich S]

On Friday, October 14, 2022 at 12:25:30 PM UTC, Don Y wrote:
[snip]

the \"standard\" american english AIUI is
standard midwestern speech, & is taught
in news journalism, and acting schools.
Siri, Alexa, et al, are based on that.
the equivalent in UK is \"BBC english.\"

so that\'s where I would start if I had to develop
a AI voice library. Or recruit an actor to
record new phoneme library. Of course
there are companies that do this already

 

[Don Y]

On 10/14/2022 12:08 PM, Rich S wrote:

On Friday, October 14, 2022 at 12:25:30 PM UTC, Don Y wrote:
[snip]

the \"standard\" american english AIUI is
standard midwestern speech, & is taught
in news journalism, and acting schools.
Siri, Alexa, et al, are based on that.
the equivalent in UK is \"BBC english.\"

But that\'s simply because someone *thought* you needed \"one
true speaking style\". Would a texan prefer to hear a generic
\"midwestern\" speaker or someone with a local twang? He\'s not
given a choice. Ditto for the New Yorker or Mainer.

Similarly for Siri, Alexa, etc.

You\'d imagine, if comprehension was an issue, all teachers,
pastors, local tv/radio personalities, doctors, etc. would
similarly be \"taught how to speak\". Pilots would abandon
Yaeger-ese, stewardesses would be taught to switch to their
\"understandable voice\" before giving directions in the event
of an emergency, phone banks would be staffed by persons
fluent in that speaking style, etc.

And, why would we let our kids grow up with the \"burden\"
of speaking a local dialect (that they would later have to
\"outgrow\" to be understood)?

so that\'s where I would start if I had to develop
a AI voice library. Or recruit an actor to
record new phoneme library. Of course
there are companies that do this already

You only \"record\" a voice if you are using limited
vocabulary synthesis (record the messages of interest)
or diphone synthesis (decompose spoken word into
phoneme *transitions* that can be reassembled to form
complete sound sequences).

[I use this form of synthesizer for people who would like to
feel they are interacting with a departed spouse or a
\"responsible older child\" (e.g., for a listener suffering
from dementia or alzheimers). Or, an ALS patient who would
be comforted by the sound of their (old) voice]

My question goes to what folks would *want* to listen to;
I doubt they\'d all pick \"midwestern US english\" if given
a choice. (if that was the case, there would be no such
thing as \"synthestic voiceS, plural\")

With a formant synthesizer, you (typically) create the sounds from
a set of cascaded filters shaping a variable excitation (voicing)
source. So, you have control over *which* sounds you make (from
pronouncing rules) as well as how you make them.

[This is a slight oversimplification]

I can, already, give the user (listener) control over various
aspects of the voice to suit his needs (some folks have hearing
and/or comprehension issues). Would you prefer a deeper voice
or one in a higher register? Fast speaker or slow? Breathy
or crisp? Monotone or prosodic?

But, I\'d not (yet) tackled the idea of altering the pronunciation
rules to favor *your* preferences for the manner in which you
expect to encounter certain words. So, those folks who like to
MAYSH their potatoes won\'t be burdened with having to map a
\"mesh\" pronunciation to their notion of \"soft, white starch piles\"

[It\'s a different sort of problem to tackle as it requires being
able to identify the things that characterize a specific \"accent\"
and goes beyond just \"changing the sound of the phoneme\" as it
also ties into prosodic computations]

 

[Rich S]

On Saturday, October 15, 2022 at 12:31:20 AM UTC, Don Y wrote:

On 10/14/2022 12:08 PM, Rich S wrote:
On Friday, October 14, 2022 at 12:25:30 PM UTC, Don Y wrote:
[snip]

[snip]

oh Don, Up top, you did not clearly express what
your goal is. Why you\'re asking. Do you want
(1) one library of speech that
most everyone will understand, to convey basic
info in short bursts (a la Siri, Alexa) [What I thought].
(2) many selectable regional dialects, for long spans
human-like conversations?
Is that what\'s needed to prevent listener fatigue?
(i.e. minimizing the person\'s mental energy demand)

 

[Don Y]

On 10/14/2022 6:13 PM, Rich S wrote:

oh Don, Up top, you did not clearly express what
your goal is. Why you\'re asking.

I hypothesized that people have biases that enhance their
comprehension of spoken word. I *assumed* that these biases
would be related to their \"dialect preferences\".

I can\'t control WHAT is said, so I can\'t use idioms that
a Mainer might use in place of phraseology that a New Yorker
might adopt (or, that a \"purist\" might express in text).

So, I am assuming that the *way* people are accustomed to
hearing things pronounced (\"accent\") is what drives this
comprehension enhancement.

Do you want
(1) one library of speech that
most everyone will understand, to convey basic
info in short bursts (a la Siri, Alexa) [What I thought].

I have different synthesizers that I deploy in different
situations.

I have a formant-based synthesizer that is best suited
for stand-alone, low resource applications. It doesn\'t
take much code or horsepower to make a formant synthesizer
that isn\'t restricted by vocabulary. (a diphone synthesizer
has a larger resource footprint -- think small, battery
powered applications like putting speech *in* an earbud)

When the appliance (earbud, in this example) can\'t connect to
anything, it still needs to be able to interact with the
wearer -- if only to tell him that there is no active connection!

Many of these messages are canned, but parameterized. E.g.,
\"Volume level 27 (of 32)\"
\"Battery level 35%\"
\"Speaking rate 250 words per minute\"
\"Battery life remaining, 3:47\"
etc.

A user will likely recognize most of these, given enough exposure
(because they are canned and, thus, predictable)

Others are sourced externally, but pronounced locally:
\"Available access points include:
greenhornet
wifi234
guestaccess
...
\"System unavailable. Contact Dr Betina at 234-5000 x234\"
\"Scheduled maintenance. Try again after 5:00pm\"
\"Account disabled. Contact Jesse at jesse@voyager.com\"

A user will likely NOT expect to encounter these and will already
be \"unhappy\" because he\'s not getting what he planned on. So,
not particularly patient with a voice he can\'t understand
speaking something that he can\'t *predict*. Having to ask for
something to be repeated (possibly many times -- what\'s the
chance that the synthesizer pronounces Betina\'s name properly?)
just adds to the frustration -- with YOUR product (even though
YOU aren\'t the problem, in this case!)

(2) many selectable regional dialects, for long spans
human-like conversations?

For prolonged use, the user interacts with a more \"resource-ful\"
synthesizer -- at the far end of the radio link.

There aren\'t really any \"long conversations\" (paraphrasing your
comment). You don\'t typically discuss politics with Siri. Or,
ask Alexa how she spent her day... (\"Oh, Siri and I went
shopping! We used your GOLD CARD and bought all new wardrobes for
ourselves...\")

So, the interactions are typically short which means they
lack a lot of context or continuity. If it is a response to
a query (or directive) on your part, then you have *some*
context in which to evaluate the REPLY.

OTOH, an unprompted utterance can be just about anything!
\"You forgot to turn off the stovetop\"
\"Your wash is ready to be transferred to the dryer\"
\"The mailman has delivered a package\"
\"It\'s getting dark. Are you sure you want to leave the garage door open?\"
\"Penny is at the front door\"
\"The freezer has warmed to a level that threatens its contents\"
\"Someone is wandering around the (fenced) back yard\"
\"The evening news broadcast is starting, soon\"
\"Time to wake up! You\'ve a doctor appointment at 9:00am\"
\"Marijane is on the phone. Would you like to speak to her?\"

These messages want to be short \"announcements\". So, you
have to make the adjustment from \"not expecting any message\"
to \"realizing a message is being issued\" and \"understanding
that message\" in short order.

And, some of them are (hopefully) very low frequency of occurrence
(the comment re: the freeezer failing, for example) so it\'s not
as if you have a past experience with it to remind you of the likely
content.

Is that what\'s needed to prevent listener fatigue?
(i.e. minimizing the person\'s mental energy demand)

I think there are two primary reasons for the fatigue.

When you listen to synthetic speech, you feel apprehensive.
There\'s a lot of anxiety as you\'re always \"on guard\" for
something that you won\'t understand (because it was
mispronounced, etc.)

\"I was scared\" might make no sense in a context where the
correct text was \"I was scarred\". So, you have to mentally
pause and think about what the correct word might have been,
instead of what you clearly heard!

Ages ago, synthesizers had all sorts of problems normalizing speech.
\"Dr. Smith lives on Smith Dr.\" \"Watch the polish maid polish the
silver.\" Or, coping with \"input errors\" (misspellings? grammatical
errors?). \"How would you pronounce\" (rest of sentence missing).
\"Dog fly airplane book\" \"1,345\" \"1,2345,456\" \"1.02.3\" \"555-12121\"

A lot of that has improved with better understanding of context.
But, even that has limitations:

(I am at the rehearsal for the play) \"I read my lines.\" (is that past
or present tense?) \"Then, the director gives me hints as to how he\'d
like me to act the part.\"

When I watch movies, I turn on the subtitles as it improves comprehension
(I\'m not having to fight sound effects, background music, etc.) -- I really
only have ONE chance to figure out what they are saying (rewinding to
replay something really impacts the presentation). Having to tell
a machine to repeat something more than RARELY quickly becomes noticeable.
Like someone who is hard-of-hearing constantly asking folks around
them to repeat what they said; it makes their situation more noticeable.

The voice (most) also has issues that make it tiring. When you watch
the evening newscast, there are likely several talking heads involved.
Often a male and female newscaster who alternate presentations -- to
give you a break from the monotony of a single presenter. Turn on
\"Narrator\" if you\'re in a Windows machine and see how long it takes before
you turn it back OFF! (close your eyes so you don\'t have visual cues
to help you sort out what it is saying!)

It\'s hard to \"program\" naturalness into speech. You can use some
general rules for prosody, breath pauses, etc. But, it still feels
like you are hearing \"Hello, my name is Bob\" repeated as if an audio
*recording* each time you encounter it.

And, you\'re listening to a disembodied voice. There are no visual
cues to pick up on (gestures), body language, etc.

I\'m just trying to make the experience the \"least unusual\" that it
can be, for \"average joes\". My design philosophy has always been to do
work so the user doesn\'t have to!

[E.g., for a user with dementia, having the voice of her eldest
child tell her to turn off the stovetop -- or go out and grab the
DELIVERED mail -- is probably more reassuring than someone like Siri
\"dictating\" to them!]

 

[Don Y]

On 10/14/2022 8:46 PM, Don Y wrote:

Many of these messages are canned, but parameterized.  E.g.,
\"Volume level 27 (of 32)\"
\"Battery level 35%\"
\"Speaking rate 250 words per minute\"
\"Battery life remaining, 3:47\"
etc.

Others are sourced externally, but pronounced locally:
\"Available access points include:
   greenhornet
   wifi234
   guestaccess
   ...
\"System unavailable.  Contact Dr Betina at 234-5000 x234\"
\"Scheduled maintenance.  Try again after 5:00pm\"
\"Account disabled.  Contact Jesse at jesse@voyager.com\"

OTOH, an unprompted utterance can be just about anything!
\"You forgot to turn off the stovetop\"
\"Your wash is ready to be transferred to the dryer\"
\"The mailman has delivered a package\"
\"It\'s getting dark.  Are you sure you want to leave the garage door open?\"
\"Penny is at the front door\"
\"The freezer has warmed to a level that threatens its contents\"
\"Someone is wandering around the (fenced) back yard\"
\"The evening news broadcast is starting, soon\"
\"Time to wake up!  You\'ve a doctor appointment at 9:00am\"
\"Marijane is on the phone.  Would you like to speak to her?\"

I\'m just trying to make the experience the \"least unusual\" that it
can be, for \"average joes\".  My design philosophy has always been to do
work so the user doesn\'t have to!

This is an online emulation of DECtalk -- probably the poster child for
formant based synthesizers.
<https://archive.org/details/dectalk>
Click the \"power\" button, then type text at the green \">\" prompt,
followed by a period and newline.

Try \"You forgot to turn off the stovetop\" (without the quotes).
Yes, you can try it again -- as I suspect you weren\'t clear
as to what it actually *said*! <grin>

Ditto \"The mailman has delivered a package\" Or, any of the above
examples. Ask yourself if you would have been able to sort out
what was being said had you NOT typed the text into the emulator!

Try words like \"police\", \"Berlin\", \"Boston\", \"Wednesday\", \"Row, row,
row your boat\" (note that it doesn\'t play out the way you\'d expect!).

There are (persistent) control codes that you can intersperse in the
input text that tweek the presentation. E.g., \"[:ra 100]\" (without
quotes but WITH brackets) will slow the speech rate to ~100 words
per minute. (unsighted users will up this to 300 or more and still
expect reasonable comprehension)

There are some predefined \"voices\" which can be selected. Use
\"[:nX]\" where X is {b,f,h,k,r,u,v,w} (the characteristics of the \"v\"
voice can be further adjusted and stored for later recall).

But, these just alter the characteristics of the \"waveform synthesis\",
not *what* is being synthesized. E.g., \"boat\" is always pronounced the
same, just \"rendered\" into a specific voice. I.e., \"Insurance\" is
always pronounced as \"inSURance\", never \"INsurance\".

Note that it is smart enough to alter the pronunciation of the definite
article (\"the\") based on the noun targeted: \"the airplane\" vs. \"the boat\".
So, emulating a speaker that doesn\'t make this alteration (e.g., B Obama)
isn\'t easily possible.

[@Dimiter, try the phrases I mentioned in my anecdote -- repeated
many times, on a single line, to simulate a continuous loop. Then,
imagine encountering them \"by surprise\"...]