I'm a court reporter (voice writer) and use DPI 15.  For years upon years, I've always added phrases to my Dragon Vocabulary to increase my recognition.  For example, ladies and gentlemen, at the time, at the time of, have you ever, etc .... 

I've always thought that there really was no limit to how much information or vocab you can add.  Is this true?  I think I read somewhere that for every word or phrase you add to Dragon, it removes one from your active vocabulary.  Is this true?  If so, how does it determine what to remove?

Just curious if adding all this is really helpful or not.

Thanks for any input you can share

I've added a small number of phrases -- 20 or 30 -- and I think the presence of the phrases improves accuracy for those phrases. However, the reason I added them was because Dragon repeatedly misrecognized the phrases, even though they appear in the writing sample Dragon analyzes when I create a new profile.

Which goes to show, that adding phrases doesn't guarantee the phrases as such being recognised, as the entire phrases that you have added, but rather the single words within the phrase.

Of course, understanding the difference is a bit above the head of the people keeping recommending to do this.

-------------------------

No need to buy if all you want to do is try ...

DragonCapture KB Download (Latest)
DragonCapture Homepage

 

Don't worry. You can always add up to 150,000 custom words to the vocabulary until Dragon starts removing existing works. A few decades ago, while I was still extravagantly open to testing things out at their extreme, I actually did the test, which wasn't so easy to set up, but it did prove true.

-------------------------

No need to buy if all you want to do is try ...

DragonCapture KB Download (Latest)
DragonCapture Homepage

I just reviewed my custom words and phrases, which I recently culled from 400-ish to 275.

About 80 are phrases. My phrases consist of two to five words. Many two-word phrases are names I often dictate, e.g., "Sandy Smith." The five-word phrase that I use the most is "an hour and a half".

Of the 80 phrases, I regularly use about 20. Dragon almost always gets them right.

It's hard to say whether Dragon correctly recognizes phrases I've added to the vocabulary that I rarely use.

Alan, you can always look up the recognition history to see whether Dragon recognised the phrase, or the single words.

-------------------------

No need to buy if all you want to do is try ...

DragonCapture KB Download (Latest)
DragonCapture Homepage

I'm a great believer in adding phrases to the vocabulary when Dragon has problems with them. As far as I can remember, I have had a 100% success rate with this technique. After defining a phrase, Dragon will preferentially use it instead of trying to take its best guess at odd combinations of words that might be in the phrase. (I smile at that sentence, because adding a phrase to the vocabulary just gives Dragon a better guess to choose; it still takes its best guess, but where phrases have been defined, the phrase is the correct guess as well.)

-------------------------

Win10/11/x64, AMD Ryzen 7 3700X/3950X, 64/128GB RAM, Dragon 15.3, SP 7 Standard, SpeechStart, Office 365, KB 2017, Dragon Capture, Samson Meteor USB Desk Mic, Amazon YUWAKAYI headset, Klim and JUKSTG earbuds with microphones, excellent Sareville Wireless Mono Headset, 3 BenQ 2560x1440 monitors, Microsoft Sculpt Keyboard and Logitech G502 awesome gaming mouse.

My experience is much like @kkkwj's - adding phrases nearly always seems to help, a lot.

-------------------------

DPG15.6 (also DPI 15.3) + KB, Sennheiser MB Pro 1 UC ML, BTD 800 dongle, Windows 10 Pro, MS Surface Book 3, Intel Core i7-1065G7 CPU @ 1.3/1.5GHz (4 cores, 8 logical, GPU=NVIDIA Quadro RTX 3000 with Max-Q Design.

I'm always getting the odd misrecognition of specific phrases that I say. As soon as I get this, and while it's fresh in my mind, I immediately say, "Add new word." Then I put the phrase in there, and it adds it immediately. Like most others, doing this helps your accuracy, especially if it is a phrase you use frequently. Also, the success rate is nearly always impeccable in most cases.

-------------------------

Tiger Feet

| DPG 15.7.1 | KnowBrainer 2020 | Windows 10 Professional /64 Bit | Intel® Core™ i9 Ten-Core Processor i9-10900K (3.7GHz) 20MB Cache |  32GB RAM. | 250GB SAMSUNG 970 EVO PLUS M.2, PCIe NVMe SSD (up to 3500MB/R, 2300MB/W) Boot Drive | 1TB SAMSUNG 970 EVO PLUS M.2, PCIe NVMe SSD (up to 3500MB/R, 3300MB/W) Storage Drive | Sennheiser D10 USB Wireless Microphone

Tiger feet, AG, KKK something, and particularly Lunis of course,

Not meaning to start a flame, but all of you don't get the overall idea of how Dragon speech recognition works.

-------------------------

No need to buy if all you want to do is try ...

DragonCapture KB Download (Latest)
DragonCapture Homepage

Can you expand on what you mean, RW? I was adding phrases to the vocab and then I moved most of them to Auto-texts, but I've added a few again to vocab lately, as I find it's slightly quicker to add something to the vocab and I get acceptable recognition.

 

Start by getting into the habit of using the recognition history to see whether or not Dragon actually recognised the phrase that you have added to the vocabulary or the single words within the phrase, or a command. There is a reason for auto-text, but definitely not when used to force a particular recognition which you should also achieve by straight dictation. Phrases in the vocabulary can help to disambiguate in certain situations, such as determinig particular spellings of proper names, as an example, or as shortcuts to produce compound signatures such as phone numbers, addresses, and the like, but they typically don't help that much within overall syntax.

 

When not used correctly, and used as a means to overcome recognition issues, auto-text and phrases added to the vocabulary are like pain killers. Rather than taking them, find the cause for the pain, and a way to eliminate it.

-------------------------

No need to buy if all you want to do is try ...

DragonCapture KB Download (Latest)
DragonCapture Homepage

Just curious - how would the recognition history of a phrase look any different from the components if there is no Spoken form? I have a couple of phrases where the Spoken form deviates from the written form - fine. But if you enter a phrase like "an hour and a half" without an explicit Spoken form, how would the recognition history be any different?

(And I know that commands are marked by an x, thank you.)

-------------------------

www.egs-vertrieb.de - www.spracherkennungscloud.de

 

Oops, I made a big mistake, and thanks Stephan for pointing it out. Of course, the conventional recognition history doesn't make any difference between single word recognition and phrase recognition. I was mixing my memory and the facts when mentioning it.

It takes some additional developing to retrieve the differences. One of my early diagnostic tools does this. For demonstration I just ran it and took a screenshot which I will paste below. For reference, prior to dictating, I added the phrase "this is a phrase" to the vocabulary, and then dictated: "these are single words and this is a phrase" into the box (without the quotes of course).

Please have a look at the results.

 

-------------------------

No need to buy if all you want to do is try ...

DragonCapture KB Download (Latest)
DragonCapture Homepage

^^^ Well, the guru was a lot more dedicated in 2012! So lower the "minimum score", the more likely/faster a phrase could be expected to be recognized?

Now is above obsolete? Using DMO, parts of which rely on DNS 9 technology, parts perhaps "avant garde", I had unexpected difficulty with calling up Toronto's "urban angel", one of its tertiary care stalwarts, St. Michael's Hospital.

On pass 1 and 2, it usually spits out "C. Michaels Hospital" or "Singh Michaels Hospital". On lesser occasions I even get "Saint Michaels Hospital".
Only upon fully articulating the word "saint" like a normal functioning "their" Majesty's loyal subject, I get my intended "St. Michael's Hospital".

But faithfully enunciating "Saint" involves combined labour of trigeminal and facial nerves and their muscle groups. I.e. too much work!

I prefer speaking "Sing" "Michaels" "Hospital".

So naturally I went to add "St. Michael's Hospital" into DMO vocabulary. It makes not a lick of difference in improving recognition odds.

Of course, I could train vocabulary to the way I pronounce or perhaps do an end-run with adding spoken form "sing michaels hospital" ... but I wasn't quite prepared to go that far.

On the other hand, for a place such as "Guelph General Hospital", I had a snowball's chance in getting DMO to output it without adding the whole phrase into DMO vocabulary. And once added with "default pronunciation", it gets put out super reliably, even if I slur.

Now before you say, Aha! That's because the word "Guelph" is not in the dictionary. But I assure you it is. Even though I can't inspect DMO's dictionary, it tells me that "Guelph" is already there if I try to add it.

Anyway, there are some idiosyncrasies, it seems.

Thanks, actually no less dedicated today than in 2012, just a little differently.



I'm afraid it's the other way round in fact. The higher the confidence scores, the faster the response. If you loosely translate "confidence scores" to "accuracy", you will find that accuracy and speed always come together and you can't have one without the other. The maximum score goes up to 999, and effectively I have never seen it manifest, but as long as the scores are somewhere near 900 or above, you will always be a happy camper. Also keep in mind that the internal confidence scores represent how confident the recogniser was that the result it came up with is what was being dictated and therefore not necessarily represent accuracy from the user point of view. For the most part, however, they correlate.



Isn't that how you should dictate all the time at best anyway? Using Dragon is like sports, you can always only expect to get good results if and when you are in good shape. So let me try now and see what I get:

St Michael's Hospital, St Michael's Hospital, St Michael's Hospital, et cetera ad libitum...

I swear that I did not train this, never dictated it before, and that it came out like that on first try. And I'm not a native speaker.

Never heard about "Guelph" before, so at first made sure that it is in my UK English vocabulary, and then looked up the pronunciation. No problem with this one either.

The reason it may be easier for you to get it correctly than with St Michael's quite obviously being down to "Guelph" not competing with so many other alternatives, even if you enunciate less clearly.

Please don't take offence, but it has always amazed me how people, specifically professionals, have such a hard time learning to use it correctly to begin with and once for all, which would definitely save them a lot of time over the course of the day, every day. In my day job as a loss adjuster, I couldn't afford wasting so much potential.

-------------------------

No need to buy if all you want to do is try ...

DragonCapture KB Download (Latest)
DragonCapture Homepage

St. Michael's Hospital

Guelph General Hospital

As you can see, both of these phrases worked fine for me without any editing or creation of phrases in my vocabulary. Unlike RW, I have heard of "Guelph" before, although I doubt that I have ever used that word in dictation.

When I tested my accuracy on a new user earlier this month, with the slider set all the way to most accurate, the Phillips SpeechMike averaged 926 and the Buddy 7G Desktop averaged 927. My actual accuracy was 98.64% for the Phillips and 98.83% for the Buddy, reading the Kennedy inaugural training text. Most of the errors were due to my misspeaking.

^^^ Thanks for chiming in some real world testing ... I had to accost a colleague today, just to find out for myself, with a little 3rd party field testing. My colleague is still on DME. For us the set-up is interchangeable with DMO.

Bad news (from my perspective) is that he had zero problem from the very get-go with "St. Michael's". Looks as though I got the monopoly on "C. Michaels" and "Singh Michaels". Some wicked "AI genie" causing mischief in my little corner of the silicon in that DMO server?

The "good news" (as in slightly reassuring) is that my colleague did muck up "Guelph General Hospital" on the first two tries. First he got "12 ... hospital" and the second I can't even remember. But he did manage to get it right on his third try!

In contrast, even after consulting Their Majesty's Manual on Royal Pronunciation, phonating out the exact "gw-el-f", as of it were some Arabic word (incidentally my colleague's first language is Arabic, though he is entirely accent-free), I managed to clinch "Guelph" without adding it to vocab ... 3 times out of 16! After a bunch of "Ralph" followed by "Golf" ...

Anyway, here is my practical conclusion:

If you are not some meticulous stickler (aka German actuarial) or aficionado American lawyer, there is no need to put your delicate speech muscles through unnecessary torture (costing you more to botox them wrinkles anyway). Just speak like a "normal", harried medic ...

When in double, add the phrase anyway. Except to get the impious AI to acknowledge a certain archangel, we might have no choice but to resort to a little spoken form, such as "sing Mikes", which is how Torontonians call that hospital anyway, as Alan would've told you.

You want AI to adapt to you. Not the other way around.

That or going to see a speech pathologist. Let me think about that.

@ax: do you pronounce "St. Michael's" in something like the way an English person would pronounce last name "St. John"? - i.e. an English person with a hoity-toity upper-class accent, where "Saint John" becomes "sinjin" (pseudo-phonetically) or "singin" when I try to put on the hoity-toity accent (which is not natural for me).

My sister went to Guelph (vet), and IIRC when I visited some of the locals pronounced "St. Michael's" almost like "Sint Michael's". Especially given how common Dutch influence is in that area.

 

I once encountered newspaper ads for "accent reduction for telephone usage" -  in the early nineteen hundreds when telephones were new. Apparently some accents just could not be heard as well on telephones.

 

I have long wondered when we will start getting accent reduction training for improved speech recognition rates.   I reckon it's a race between speech recognition becoming popular and machine learning being able to handle more and more accents. I would probably bet on the ML.   But in the meantime,  I consider it not all that bad that it forces me to enunciate more carefully.

 

Now if I could just get Dragon to not map "git" to "get".  

 

-------------------------

DPG15.6 (also DPI 15.3) + KB, Sennheiser MB Pro 1 UC ML, BTD 800 dongle, Windows 10 Pro, MS Surface Book 3, Intel Core i7-1065G7 CPU @ 1.3/1.5GHz (4 cores, 8 logical, GPU=NVIDIA Quadro RTX 3000 with Max-Q Design.

^^^Your sister graduated from that one-of-a-kind OVC? Up until now I thought you were the smart one in the family ... (never mind what RW thinks).

"Hoity-toity"? Absolut nicht! My tongue wags closer to the pidgin/patois end of the spectrum. Not "les patois Québécois", however.

Nevertheless, you just gave me the clue that fixed my "St. Michael's" problem.

My problem wasn't that I preferred to say "sin- or sing-" as opposed to "sain-". Mine was a refusal to even acknowledge the "/t/". I literally say "sin/sing-michaels" without pause, which begets me either "C. Michaels" or "Singh Michaels".

I was practicing "/t/ elimination", whereby the proper rendition should be "/t/ glottalization", ie, the "/t/" goes to a "glottal stop". Thus it doesn't really matter whether one says "sin-/sing/sane-" so long as it is followed by a staccato pause before saying Michaels. With that DMO puts out "St. Michael's" for me without fail. As an aside, if the "glottal stop/pause" is just a smidgeon too long, then DMO puts out "Saint Michaels Hospital" (sic).

Bloody /t/ ... and bloody AI.  Funny thing is, Dragon doesn't give one whit about the English "th" sound, which is not in French or German.  "Thousand" and "Sauzand", "zis" and "this" give you exactly the same thing.  But it chooses to hang up on that little "glottal stop".  I am guessing it may in essence be a word boundary thing whereby it picks up on teeny differences in timing.

Back to "Guelph". The middle "vowel team" is highly diphthongized. Just can't cut corners on that one, which meticulous Germans (who are used to working their facial muscles - botox money be d@mned) and well-spoken lawyers generally don't. So they tend to be fine.

Too late for me to study to be German lawyer.

 

Tsk, tsk. The first part is simply reactionary, unfit for Anthony Blinken's revolutionary New America. The second part gets us closer to the zeitgeist. Leon once said: "If the people cannot follow the Vanguard ... git me a 'New People'!"

If English cannot work for the "people", then git me "People's English", through ML, or vocabulary editor.

In real life, I am all for compromise. Therefore I took your excellent hint on "sint", and with just the right little "glottal pause" between "Sing" and "Michaels", I now get "St. Michael's" reliably.

But I am not putting up with no diphthong or other thong. I added "Guelph General" to vocab and Q.E.D.

P.S., watch BBC and learn the language, Ag!

AX,

Your reflections posted continuously to the forum never cease to amaze me, and now for some of the details:

Certainly too late for you on more counts than this, but here's a little test I would suggest for you to go through. Are you able to get Dragon to distinguish between "dwell" and "Dell" reliably, which equates to the same delta of involving "facial muscles" and overall articulation impact? And if you do, you should also be able to get "Guelph" distinguished from other things such as "give", "give", "golf", (which I have just pronounced deliberately without diphthongs) without any additional fussing about.

All of such postings keep making me mourn the days when people starting with Dragon were forced to go through the initial reading which quite clearly stated that Dragon listening to speech is entirely different than humans communicating, and at best, when talking to Dragon, you should always make sure to sound like a newscaster reading the news.

To that end, if you insist on ignoring diphthongs being useful in spoken conversation, consider yourself lucky if you manage making yourself understood in your hood, but don't count on getting the concept across in other situations, such as when talking to people on the phone who are not familiar with the background, as an example, assuming that you may be in a situation like this.

Communicating is all about being clear. Unless you're doing arts.

-------------------------

No need to buy if all you want to do is try ...

DragonCapture KB Download (Latest)
DragonCapture Homepage

Can't argue with the desirability of articulating clearly.  Even minus the prospect of a German legal career, that's something I owe to my current "clients" and to my profession.  I do have to remind myself to speak slowly and not let the facial muscle slack off too much ...

In fact Lindsay's long-held advice on recording oneself once in a while through Windows Voice Recorder is still golden.  Not just vis-a-vis Dragon, but also in self identification/correction of poor pronunciation habits.

And I do wish to thank yourself and others for taking the time to test out "St. Michael's", as simple as it may have been ... It made me realize where my problem was - shortening the "Long-A" is no foul.  But skipping the "T-pause" is no-no.

As to "dwell" or "Dell", these are practically unrealizable in the context of DMO's medical vocabulary - at least for me.  However, upon switching to the "Clinical Administration" vocabulary, and especially used in phrases such as "Dell Computer", there is no problem at all.  Specifically, "dwell" is a bit interesting.  Even when used in full sentences, it stubbornly comes out as "do well" for me in all sorts of combinations, with the medical vocabulary.  Switched to "Clinical Administration", it only comes out super reliably when used idiomatically as a prepositional verb, following some sort of infinitive, as in "don't/not/to dwell on xxxx" ...

DMO's "AI" language model does appear to display some "natural language" awareness.

Anyhow, I was semi-"tongue-in-cheek", literally and figuratively, in my dismissal of "diphthong".  If we care to split the peas on "Guelph", I would have to first stand corrected as there is no "vowel team".  There is a "consonant cluster", where "u" takes the place of "w", by some archaic Latin pronunciation convention, followed by a diphthong, ending with a digraph.  Does knowing any of those jargons help anyone?  Overall it's a relatively obscure word, producing a relatively awkward syllable.  Once I determined that it is comparatively error-prone in DMO, whether spoken by myself or by colleagues, it is not hard to conclude that it's simply not a "hill worth dying on".

Btw, another phrase I recently had difficulty with was "patient delegate", in which the second word rhymes with "delicate".  Invariably and perhaps not surprisingly, DMO insists on putting out "patient delicate".  I consider that a "natural language" fail. 

 

The simplest workaround for me is to pronounce "delegate" in its verb form, which rhymes with "dedicate".  Except to the "purist" this is nothing short of grammatical butchery.  Frankly, I haven't got the time to be a language sentimentalist.

P.S., what I am about to say may seem "unfair" to the English tongue.  But it's difficult to view English as "just another language".  English has a mission, for better or worse, different from that of Deutsch or Français.  If a large swath of its speakers could dispense with "rhoticity", for various reasons, is it such a sacrilege to re-evaluate the worthiness of certain "diphthongs" from time to time?

I myself would rue the day when a dodgy phrase such as "force majeure" breaks into popular vernacular.  Are we going to sweat over a Latin "triphthong" now if using a non-lawyer version of Dragon?  Any wonder some prefer to take it out with a spoken form ... with a show of "force major"?

Not sure I add anything not already contemplated but one thing I do notice is that mucking around adding phrases and pronunciations as a way of getting Dragon to respond to a particular articulation (what I call an utterance) is a losing proposition in the long run. Find out how to get Dragon to respond natively and train yourself (I bet you can learn it faster than Dragon). When it comes to sloppy or tired utterances, resist trying to turn them into phrases as a shortcut. Actual shortcuts, though, where the phrase has a true alternative spoken form, are of course to be utilized.

What do you do with a phrase like "time period" which if you say time. Without the quotes you get something strange. Here is my recognition history:

what do you do with a phrase like open quote time period close quote
which if you say time period without the quotes
you get something strange period

Of course, "time period" without the quotes has been a phrase in my vocabulary forever, and under DNS13 it was dependable. Under DPG15 just now it didn't work the second time (even though I used the phrase in the context of the quoted expression just before) but after correcting it once it became dependable to say time period even at the end of an utterance and get it right. Here is the recognition history for that:

dependable to say time period
even at the end of an utterance and get it right period

So my advice is to start with a new unadulterated profile and do as little as you have to with it. If you have it, the data distribution tool can help you keep pristine lists of custom words and commands that _are_ useful and using different vocabularies can help you segregate idiosyncratic problems from infecting the rest of your general purpose dictations.

My 2¢

-------------------------

PG

Remember folks, my comments and this forum are for entertainment value only, please, no wagering or other reliance on the contents herein.  I permit no commercial use of my ideas (whether expressions or embodiments) without my written consent.

PG,

I haven't read truer words in any posting on this forum for decades, or so it seems. I couldn't agree more. It is worth not just $0.02, but probably $2 million. No kidding on my end this time I swear.

-------------------------

No need to buy if all you want to do is try ...

DragonCapture KB Download (Latest)
DragonCapture Homepage

I use 'period' to get a '.' and say 'period period' to get 'period' in the buffer.

-------------------------

Win10/11/x64, AMD Ryzen 7 3700X/3950X, 64/128GB RAM, Dragon 15.3, SP 7 Standard, SpeechStart, Office 365, KB 2017, Dragon Capture, Samson Meteor USB Desk Mic, Amazon YUWAKAYI headset, Klim and JUKSTG earbuds with microphones, excellent Sareville Wireless Mono Headset, 3 BenQ 2560x1440 monitors, Microsoft Sculpt Keyboard and Logitech G502 awesome gaming mouse.

To add to the above contributions.  If you want a further example, perhaps easier to demonstrate, why I would never add phrases to the vocabulary and never have had the need to, other than the proviso "where the phrase has a true alternative spoken form" try the following:

1) Add the following phrase to the vocabulary "So I subsequently vetted my emails"

2) Then dictate the following "So I subsequently edited my emails"

The findings with my profile at least are when dictating the phrase "So I subsequently edited my emails" Dragon has a tendency to skew towards the phrase that has been added in 1) above rather than what I actually said and thus the outcome is incorrect.  Appearing as if Dragon looks for a phrase match from the vocabulary (getting a suitable "score") before doing what it should be doing…

 

So further example of why it messes up your profile, I would never have problems with either phrase being recognised 100 percent of the time until adding the phrase.

-------------------------

Whether intended or not, this is another hammer that hit the nail on its head. First of all, if I didn't have problem with "either phrase being recognised 100 percent", I would hardly bother adding either to vocabulary in the first place.

This "skewing" effect is exactly what I am looking for. If by adding the phrase "Guelph General Hospital", the recognition odds of the alternative "12 xxxx hospital" or "Golf general hospital" are lowered, Then I'd call that a feature, NOT a bug.

Above may be true for Desktop Dragon. I draw a different conclusion concerning DMO/DPA, based on first-hand observation so far. For one thing, I notice that using the corrections menu has no discernable effect on improving subsequent recognition odds, which is a bit of a bummer.

In fact, I would love nothing more than to be able to "muck up" the DMO vocabulary and see how far I can take things. Except I am prim and proper to the point of being prudish, in large part because I can't afford to divorce my present "profile" for fear of losing my precious AutoTexts and step-by-step commands (50% invoking AHK - but just putting in the triggers still took a valiant chunk of my time), while I have no attachment to my current vocabulary per se.

The reason I couldn't "muck" with the Clinical Administration Profile is because it shares exactly the same (one and only) custom vocabulary with my Medical Profile in DMO. Perhaps desktop Dragon 16 will finally let me have proper control for once and solve the "mucking" problem.

Regrettably there isn't any "data distribution tool" I know of. Does anyone know of such a thing for DMO/DPA? Technically I am still "testing" DMO ...

Anyhow, here is a question to any informed DMO/DPA VAR who may come across this thread: how do I ensure "bulletproof" transferability of AutoTexts and Step-by-step commands during profile "refreshment" in DMO? Should that be "automatic" or there is a "defined process"? All I know is that I suffered no AutoText loss during switch from DME. But other colleagues weren't so fortunate. Anyway, this is a digression. Any knowledge on that is appreciated.

 

I didn't add it as a phrase, like I said in the post I don't add any phrases to the vocabulary (other than "replacements") and so this was posted purely as an example.

 

 

 

I will leave that up to you

-------------------------

Now let's expand.

Far be it from me to impugn the general sagacity of the Tier 1 Dragon experts who weighed in on this subject. But the experts are not necessarily in the foxholes. With me.

What do I mean by that? Well, do Dragon experts dictate on busy wards, ER, OR where bells ring, alarms ding, people chat, patients groan, and metals bang? Even RW, by his own historical divulgement, dictates in more hospitable, controlled environs. If vocabulary methods give one just the right little "nudge", or "skewing" as in Lindsay's parlance, and help one survive the trenches till the morning, then is it not a valid tactic? Just know that it is no free lunch and can have "blow-back".

Using another medical jargon: "Phrase addition" may have a narrow "therapeutic index" ... according to ChatGPT.

Was it wrong for the Junker wigs to teach their Prussian foot soldiers that "aim well, and shoot straight" beats "wasting bullets" at all times? That the Mauser was the pinnacle of bolt action and bolt action was all they ever needed (according to ChatGPT)? And "they be right" ... until 1943 or thereabout. The wisdom didn't expire. Just that the battle changed.

Back to our peaceable endeavours, if a couple of colleagues using the same general flavour of Cloud Dragon all have the same difficulty with a particular phrase/word, then by all means - invoke the vocabulary methods!

It's one thing to teach as the Kommandant of Dragon-Lehr-Regiment. It's quite another to do what you must to survive the foxhole ... till the morning.

Now let me summarize:

1. Don't be lazy. Pronounce well. Exercise those facial muscles (more of which I do now, after my "St. Michael's" debacle).
2. Absorb the sage advices (contrary to any false impression, I do so in spades)
3. Ask questions
4. Synthesize the information. But do what you have to anyway.
5. But at the end of the day, any serious subject deserves to be approached with discipline and study. That might be the real advice from RW et al, which, unfortunately, does not go out of style, AI or no AI.

P.S., now back to beating my new favourite "dead horse", the diphthongs. The last syllable in "nuclear" is one of the 7 well-known English diphthongs, which rhymes with "ear". However, DMO will perfectly accept "Nookular Medicine". Who says I need to speak English? I just need to speak American.

P.S. #2, which segues into the subject of "People's English". Far more than that, I am hoping that with ML, soon Dragon will be able to understand post-stroke dysarthric patients' enunciation better than humans. This is among the real benefits and progress, in my mind. After all, is Dragon fundamentally not an assistive technology?

In RW's example, he definitely picked two very troublesome words. I have consistent issues with 'edited', 'edit it', 'ed id' because I don't enunciate the 'ed' and 'it' clearly enough (I get lazy when I'm going at speed, or else I learned to pronounce those words that way. Ditto for 'vetted', 'betted', 'edited' (although edited has three syllables and some rhythm). My approach has been to define key phrases where I have to use those words, add blank audio space around them to isolate them a bit better for Dragon (no slurring them together), or just use different words entirely.

-------------------------

Win10/11/x64, AMD Ryzen 7 3700X/3950X, 64/128GB RAM, Dragon 15.3, SP 7 Standard, SpeechStart, Office 365, KB 2017, Dragon Capture, Samson Meteor USB Desk Mic, Amazon YUWAKAYI headset, Klim and JUKSTG earbuds with microphones, excellent Sareville Wireless Mono Headset, 3 BenQ 2560x1440 monitors, Microsoft Sculpt Keyboard and Logitech G502 awesome gaming mouse.

 

This is against the rules at Dragon University, also referred to as "cheating", and one of many ways to screw the exam.

 

For any combination of pairs like "xyzed vs. xyz it", just insert an ever so slight pause before the "it".

-------------------------

No need to buy if all you want to do is try ...

DragonCapture KB Download (Latest)
DragonCapture Homepage

 

The example I posted with edited and vetted was not to highlight or distinguish between those 2 words(I don't have problems with that) but just purely as an example of what adding phrases can do to your vocabulary in terms of skewing. I could have equally used different words to highlight the same issue.

-------------------------