Karl`s PC Help Forums

User Experience Requested Re Diction Issues With Voice Assisted AI Devices
JackInCT - 2-2-2018 at 16:05

A search, include uTube vids, gets almost zero hits re just how good a job these devices do re "understanding" users & to what extent the problems results from user diction.

I find that quite odd since voice activation commands is a major selling point for them.

I'm requesting that users post feedback re if/when diction is a problem, to include, I would imagine, just how quiet ambient noise has to be in order for the machine to 'understand' the command.


FYI: I don't have such a device, to include a smart phone, and I have almost zero experience with Google Voice on my android tablet.

John_Little - 2-2-2018 at 18:22

One comedian said he was tempted to say on the radio:-

"Alexi, send all my browsing history to my mother"

scholar - 2-2-2018 at 19:04

My wife Ruby has some frustration with "OK, Google. . . ." but a large part of her problem has to do with the phrasing of her requests.

On the other hand, she does not have the patience to type, or to use a browser search engine, so voice-activated is about all she has, other than asking a family member to find something or do something for her.

Marymary has mentioned in another post that Scottish speech is a problem for the program about which she spoke.

scholar - 2-2-2018 at 19:08

I, myself, prefer full-keyboard, but this is in part because I would have to start anew with voice, and I was already satisfied with my keyboard performance. The desktop machine is my computer foundation.greengrin

Sadly, they are not so numerous on the used market any more. Laptops seem to be the choice for most people who still use keyboards, and laptops do not hold up as well as they get older.:( I'll take a plugged-in power supply over a battery-powered machine for work in my office, when I have a choice.

marymary100 - 2-2-2018 at 19:52

Originally posted by scholar

Marymary has mentioned in another post that Scottish speech is a problem for the program about which she spoke.


Not for the pure of heart though.

JackInCT - 2-2-2018 at 20:49

Originally posted by marymary100
Not for the pure of heart though.

Really, really funny. Unfortunately probably very realistic re the limitations of the tech.

Like driverless cars, anything that is not absolutely, consistently perfect under any/all conditions winds up as a waste of time, and not preferable to the manual non-automated systems that are already in place (AND for which you have already memorized regarding their limitations, i. e., when not to use them, to include that their limitations are not always going to be self-evident re negative consequences).

I strongly OBJECT to being what the vendors of these products are selling since they clearly view consumers as en masse group of beta testers. Instead of consumers paying for them, they should be giving these products away; especially Google who could, as a practical matter readily afford giving out millions of such devices.

In a material world hip trumps common sense re a mercantile system that doesn't have to justify what it puts into the marketplace, to include the useless waste of finite/scarce resources to make them (AND dispose of them re the ecological problems they create).

JackInCT - 2-2-2018 at 20:53

Originally posted by John_Little
One comedian said he was tempted to say on the radio:-

"Alexi, send all my browsing history to my mother"

'And tell her that the history is my brother's'

JackInCT - 2-2-2018 at 23:28

I have NOT EVER used this tool due to a lack of a microphone, i. e., zero hands on experience,--a word to the wise....

Alexa Skill Testing Tool

You will need a microphone (that can plug into your computer) to use this service.

"The Amazon Echo Simulator (Echosim.io) is a browser-based interface to Alexa, the cloud-based voice assistant that powers the Echo. Echosim.io is intended to allow developers who are working with the Alexa Skills Kit (ASK) to test skills in development. While Echosim.io allows you to interact with Alexa, it is not a substitute for Alexa-enabled devices, like the Amazon Echo. For example, Echosim.io does not stream music or support far-field voice recognition."

"An Amazon account is required. Log in with Amazon."

Echosim.io Community Edition by iQuarius Media

Me Here: no idea who owns this site. No specifics at the URL re actual owner.
No forum re user comments.

Suspected caveats re obtaining optimum results:
(1) NOT all microphones should be considered "equal" re input [MOBO mic in versus a dedicated sound card mic in should be considered],
(2) No idea if CPU speed could be a factor with results [shut all programs down],
(3) No idea where this server is, i. e., if results diminish with geographical distance from you to server,
(4) No idea if DSL speed is a factor,
(5) No idea if number of users logged in at any one time is a factor, i. e., the more there are, the more the server gets "bogged" down,
(6) No idea what effect ambient nose level has, i. e., probably should be tested in a quiet area,
(7) No idea if logged in to the site on a smart phone what impact that has versus a desktop/laptop PC,
(8) If possible, 2 people giving the exact same command back to back and noticing if results differ might have value,
(9) If you have more than one machine, maybe a comparison with same commands would indicate if that's a variable; ditto re OSs.

LSemmens - 3-2-2018 at 00:29

I loved the Scottish elevator!:D

I worked with an early voice dictation system back in the 90s. A customer wanted a demo and, as our technician was not there it fell to me to demonstrate it. Given that in those days it had to be "trained" to understand your voice and our tech was the only one who had done so, it worked surprisingly well.

Modern stuff is streets ahead. The thing I hate about it is that businesses seem to think that a machine is a good substitute for a receptionist, or worse, for making telephone calls!

JackInCT - 3-2-2018 at 02:29

Just for the sake of a simple experiment, I dug out my 2 yr old Samsung Tablet [Galaxy Tab S2-Anroid Ver 7.0-All available Android updates installed] and tried
Google Voice with a DSL wifi connection.

First I asked for a list of computer help forums, and this forum was nowhere on the various hits.

I then asked for Karl's Forum, and the first 3-4 tries with variations in the name got none, BUT then one last try got the URL. That's unacceptable in the real world, i. e., no one is likely to try and try again for a hit. Rather, it would be simply easier IMO to go to the tablet's browser and search there and simply reword by typing the search words--it's not a speed issue, i. e., the hits were very quick in coming, it's a success issue based in part I would think on the user's choice of search words, AND diction, i. e., I could see as the suggestions came up that for a while it was Carl's forums Google search for and not something like "Carl, or Karl, forums". I'm not blaming Google; when all is said and done it is still immature tech, and immature tech = time wasted, especially when alternative means exist that seem slower but actually over the long run are more efficient. I still however hold to the possibility that at least a part, and perhaps the core issue is diction/accent/AND perhaps the quality of the mic in the tablet. That is why in my above post re the Alexa site I suggested that the various experiments involve 2 or more people to see if different "voices" using the same "request" get the same, OR different results.

LSemmens - 3-2-2018 at 08:45

I'm with you there, Jack. Personally, I find it quicker to talk to a human, otherwise, I'd rather "talk" by typing on my computer. I HATE text messaging and use it only as an evil necessity.

JackInCT - 4-2-2018 at 00:07

Originally posted by LSemmens
.. Personally, I find it quicker to talk to a human, otherwise, I'd rather "talk" by typing on my computer....

If this has NOT already occurred to you, allow me to point this out.

In some quarters (and most certainly NOT necessarily mine) Alexa is perceived as a shill for Bezos re his accurate perception that homo sapiens have, essentially, an unbridled lust for material possessions, i. e., they like to spend their money for the sake of spending it re an outcome that it makes them feel sooo good to do so.

One of the major capabilities of Alexa is to purchase a selection of items on Amazon [which it calls "Prime-eligible physical products"] by talking to Alexa [you cannot buy everything that Amazon sells (you would have to revert to using your browser if you wanted somethingthat is not on the Prime list), and you need an Amazon Prime account, and some other requirements/limitations]--in the trade these are known as impulse purchases.

I doubt if the first version of Alexa was released until this capability was available.

If you can talk, you can buy.

And when delivery drones take to the air...

And, as Plan B, if you live in an area where its same day delivery service is available,

And as Plan B, Subpart A, if you wish to let the delivery person enter your premises when you're not home....

Tell me why I am typing this reply when I could be spending my money like there was no tomorrow?????

So as an inducement, let me give you this scenario: "Alexa, I want a keg of [fill in the blank]; which distributor can deliver one to me the quickest".

JackInCT - 4-2-2018 at 03:59

FYI: I spent about an hour testing a variety of phrases on my adroid tablet "ask Google"; this is NOT intended to be a full fledge comparison review.

Some of the responses were voice, & some were text. Voice responses were crystal clear/perfect tonation.

"Easy' questions such as what's the time, what's the date, what's the weather were voice. They were all instantly answered.

(1) If just one word is 'misheard' [especially a noun] the entire search is wasted,

(2) Even when the entire phrase is heard correctly, the more obscure (however subjective is the user impression of 'obscurity'), the odds are great that there will be NO hit for the answer,

(3) Some very long phrases can be spoken and Google will search the entire phrase; it was difficult to quantify if speaking a long phrase slowly helps Google to hear the entire phrase; it was also very evident to precisely determine how long of a pause Google will 'tolerate' before it 'decides' that you are finished and begins a search--NONE! Breadth control is critical as the length of the phrase increases,

(4) I did NOT incur a single instance where Google stated that it did not understand the search question. When it obviously didn't, I got junk hits.

IF there are settings in Google voice to 'adjust' how it works, I did not look for them. On my todo list.

Very impressed with the ability to answer correctly easy questions. Even a somewhat complicated question of, 'how far from my house to San Francisco', the voice told me that I needed to turn on a certain feature before it could answer it.

So from an enclopedia perspective of answering questions, Google voice seems a possible worthwhile alternative to Alexa. I have no home automation devices; so I cannot test if the tablet can download an app for the device, and whether Google voice can activate the device when I tell it to, i. e., it's worth exploring if an existing device is a worthwhile substitute for an Alexa type additional device.

It seems conceivable that the major difference between Alexa type AI devices, and a smart phone is that the user is tethered to the phone to get it to 'perform', while Alexa can be positioned just about anywhere in the vicinity (don't know if there is a battery powered 'mobile' model). The speaker quality can be a major factor if the device is used to play music and you choose not to use it's line out to your existing home audio gear. But then there's the blue tooth alternative for phones to connect to blue tooth enabled speakers.

JackInCT - 2-3-2018 at 20:30

Originally posted by JackInCT
Alexa Skill Testing Tool

Feedback/Review of my initial visit to this site.

(1) I have zero idea if this site's Alexa (and its answers) is identical to the actual device's answers,

(2) If you're not use to having a microphone on your PC, a test, before you begin, is in order to confirm that your gear works, to wit: if your microphone is connected to a headset, AND if you have speakers plugged into your line out port on your PC, you may need to have the speakers powered on to hear Alexa [in Windows OSs, there is Recorder under Accessories and that would be a good place to start re testing the mic, and hearing itplayback],

(3) This site does NOT support all browsers, plus you may have to deal with cookies on, and pop-up blockers settings, to get it to work,

(4) You definitely have to log on to Amazon to get this site to work, i. e., no Amazon account & you're out of luck,

(5) There is no transcript available of your "conversation" apparently, i. e., you will have to trust to memory for answers; that's more than inconvenient since as a 'test', when you don't get an answer, you will likely try rephrasing your question to see if that works,

(6) There is not much of a delay between question and answer, but obviously that will depend on CPU, modem speed, etc.,

(7) Alexa does a great job of speech recognition, but you really should clear your throat, annunciate slowly & clearly, etc., to include minimizing background noise,

(8) I have an Echo Dot on order; an unboxing review in about a week,

(9) The more concrete the question, versus any level of abstraction, gets the best correct answers, but how many times a day are you going to ask 'what time is it',

(10) Even a concrete question such as the distance between NYC and San Francisco which gets the correct distance, falls flat on its face when you ask how many inches there is between the 2 cities,

(11) A test of this site is IMO the equivalent of talking to a toddler; IT IS A NOVELTY/rainy day amusement pastime item as I'm quite sure the Dot will turn out to be,

(12) I also ordered a Smart Home light fixture adapter, and that area, re my prediction, will be Alexa's strong suit, i. e., I expect that with the Dot's Far Flung microphone system I can, as an example, hook up the adapter to some location where I might go in the dead of night, and have the light, upon a verbal command, come on to show the way [FYI: these adapters are not cheap],

(12A) There are of course many positive possibilities for Alexa for folks as a form of adaptive equipment; my go-to tech person mentioned that an elderly family member who lives alone, & who has a significant visual impairment, has Alexa available for what amounts to a verbal information source similar but not identical to what is on the radio; there are Alexa apps and it would be very interesting to see what Alexa can do that can improve the quality of life for people with varying degrees of disability,

(13) As far as information retrieval goes, you can get the same info via a smartphone such as with Google Assistant; I've not tried to set up my Android tablet to see if there is a Google Assistant app that will verbally give me that info via the tablet's speaker; ditto re the Smart Home gear; BUT with the zillions of apps on the Google Playstore, at the very least, you're in a time bind re 'discovering' if any given app really does deliver as promised re trial & error. With Alexa, the learning curve is likely to be much shorter and definitive one way or the other.

JackInCT - 10-3-2018 at 21:57

Well I took delivery of my Amazon Echo Dot (AKA Alexa) yesterday, and there were a few surprises.

To begin with since it appears that no one else on this board has one, in order to work the Dot, you first have to download the Alexa Skill on your smartphone/tablet [FYI: everyone else would call Alexa an app, but Amazon decided to use some different terminology]. As a beginner, I think that the download CANNOT be done via a PC, etc., but who knows?

The skill is where you set up Alexa, to include the all important connection to a WiFi network which in turn connects you to Thee Cloud (at the level of a guess, I think that it's safe to presume that the Cloud has a 24/7 direct connection to some financial institution).

The purpose of this post is 'narrow' re an FYI as to how one aspect of Alexa "works" in the real world.

To wit, all questions that you put to Alexa, ANDfor which 'she' has an answer:

(1) no option for any variety of a voice selection, except for a given country selection, i. e., such as the UK option gives you an English accented female voice, BUT it impacts how Alexa answers questions (see next paragraph),

(2) it will answer questions presuming that you live in the UK, and "phrases" the answers upon a presumed UK culture--that's right, the same question asked in the USA versus the UK can get you a different answer,

(3) for questions that it has an answer for (versus ones that either it ignores [and that happens a good deal presumably because the question is toocomplex for it] to include an answer of "I don't know that"], the question and the answer immediately come up on the Alexa skill [if you have itopen], i. e., the questions that you ask that it ignores, or can't answer are NOT "logged in".

Bottom line: Big Brother has a record of what you ask, and the answers that Alexa 'generates'.

The embedded image is but a small sample of all the questions that have turned up in the skill since I installed it.

What's it doing when it's just sitting there? And yes, it's internal "far flung" microphone is really, really good at a goodly distance away from it, relative to the level of background noise, and ditto re SUPERB voice recognition to include if you ask a question with a mouth full of food.

BE VERY CAREFUL WHAT YOU WISH FOR! aka 'don't be stupid and get caught'.

JackInCT - 11-3-2018 at 01:56

FYI Addendum: As it always seems to happen with new tech, 'tricks of the trade' occur on a happenstance basis re some hit verbiage.

I've discovered that I can view, apparently in its entirety, the Alexa skill that I have on my Android tablet, on my desktop PC (which uses an Ethernet connection).

The URL for this is, https://alexa.amazon.com/spa/index.html#cards

Obviously, you will need to sign in to you Amazon account to use this URL, as well as have a Dot, i. e., I presume that if you click on the URL without BOTH, you will hit a dead end.

The sole purpose of this 'capability' (so far at least) is that if you had, as a for instance, a grocery store list skill, you could do a c & p into a word processing type file, and then do a printout to take to the store. I haven't tried a grocery store skill yet, but I presume that you could simply verbally tell Alexa to add items to your grocery list, then c & p it on your PC, etc.,. And I presume that, POSSIBLY, you could blue tooth that file to your smart phone and use that at the store rather than a printout.

JackInCT - 11-3-2018 at 03:24

FYI: Addendum #2

There are free skills that create grocery lists, and which will send said lists to your smartphone. They all have varying degrees of very low negative reviews.

The native Alexa skill has a "list" setting, and I tried it for the first time re compiling a grocery list and while not full of bells and whistles, does the job (less sending it all on its own to a smart phone); each item has to be entered one at a time; I suppose you could have a Dot in the kitchen area and add items on as needed basis. That search on the far right is hot wired to Amazon, i. e., each item can be bought on Amazon to include the hamburger--I Kid Thee NOT. It is shipped, and apparently doesn't spoil in transit--lots of luck with that, i. e., hamburger is the "cut" of beef most likely to be infected with a pathogen [my "GUESS" is that it is shipped after being irradiated].

The embedded image is the first several items that I could do a screen capture of.

LSemmens - 11-3-2018 at 08:08

Maybe we should merge this thread mwahahahahaaaaaaa!

JackInCT - 12-3-2018 at 01:35

FYI # Whatever:

IF you play a utube vid re demo Alexa's features, Alexa 'thinks' that's "you" talking and will comply with whatever 'command' the person doing the vid makes.

That of course leaves you with all sorts of reminders/alarms/etc. that you have to then figure out how to remove [which if you don't know how to do that, leaves you with a new chore to learn, especially if the demo says, set an alarm for 3:00 AM].

By the way, the Amazon technical nomenclature for the word "Alexa" is "wake word"; it can be changed in the Alexa app to a few other choices for a wake word. What are the odds that someone named "Alexa" will EVER come to visit you? Owners, especially those that use the Dot for music, often connect the Dot to either a Bluetooth speaker(s), or their PC's stand alone speakers; some of the more expensive Echos have better sounding speakers.