Correct pronunciation with Dragon Naturally Speaking Transcription Software.

[mentecuerpo]

I have had excellent results with speech-to-text using Dragon Natural Speaking, Google Translate and other online STT applications with poor microphones and good microphones. I have a very nice Yeti microphone that is excellent for Skype language chats because it is high quality, and it has its own audio out (that lets you listen to your own voice and the incoming voice without any latency issues). But using the Yeti isn't necessary for good quality STT. I think the main ingredients of successful STT are clear pronunciation and careful, deliberate speech, not the quality of the microphone.

Thank you for your contribution.
I know you speak Dutch Tommus, I have read some of your interesting posts.

It is important what you mention here because I don't want other people to think that this tecnology does not work and prevent them to give it a try and check it out for themselves.

I have even played an Assimil German lesson while have a google doc open with the "Voice Typing" feature, and I get a very good transcription. Granted, no punctuations but the text is accurate. I prefer to use Drago that google, I believe the two programs work differently.

In the hospital I work, I use Dragon so many other workers there. I have noticed that the transcription errors are not errors but me not knowing the correct pronunciation in English.

[mentecuerpo]

My understanding is that transcription software adapts to the speaker, rather than requiring the speaker to adapt to its model of language.

That sounds like a bit of a problem...

You can train it, but usually the software likes standard accents. For example, when I want to say, "the patient discharge" Dragon will transcribe: "the patient this church" and its because my Spanish accent and my intonation, so I need to be very careful to say "discharge" correctly, otherwise, Dragon thinks I said the "church" thing.

[Cainntear]

Also, Windows has inbuilt speech recognition -- is Dragon sufficiently better for this that it's worth paying extra?

[mentecuerpo]

Also, Windows has inbuilt speech recognition -- is Dragon sufficiently better for this that it's worth paying extra?

Yes, the accuracy changes, Windows is not good at all.

You can try a seven day trial with "Drago Anywhere" it works in Android and iPhones. You have three languages you can use, English American or UK, and German.

Try with your English native accent, and you will probably get above 95 % accuracy.

Then, try with a few phrases in German.

First, play a few sentences from audio with text to have a control group. Turn the microphone on and play the audio and let the app transcribe the sound; it should be 95 % accuracy.

Then you will try to pronounce the same sentences and let Dragon Anywhere transcribe the text. You will have transcription errors until you can reproduce the phonemes, then you will get the text compared to the control group.

I did some Glossika today, with German. I would listen to the German phrase in Glossika, and then I would try to replicate the sound after and Dragon Anywhere will give me immediate feedback by the from of a transcription in text; the wrong transcription means an incorrect pronunciation.

Theoretically speaking, if I read a phrase and it is transcribed correctly by Dragon, then a German speaker should understand the sounds I am making.

As I got better at it, and the transcription from Dragon Anywhere matched the text from GLossika. Except for a few words that I can not replicate the sound.

I will need a German teacher to help me with the words I cannot pronounce correctly after multiple efforts. Pretty cool.

I was able to transcribe a German youtube from the "Easy German" series. The transcription was almost perfect.
It works better if you extract the audio, then play and pause at different phrases.

The idea is to let the software send the audio data to the dragon server and receive the processed data from the server. Dragon anywhere requires an internet connection or your cellular plan. However, the PC software does not require an internet connection.

Dragon anywhere is expensive, $15 a month, but I will use it for the 7-day trial. After the seven days, I will cancel the Dragon Anywhere trial, but I have ordered the PC software that is not subscription base.

I bought Dragon Naturally Speaking Home edition, German version (it comes with German and English) via eBay from Germany.

[mentecuerpo]

Speech Recognition Software to learn the correct pronunciation of foreign languages:

I am incorporating SRS to my pronunciation drills. I am pleased with the results and the instant feedback I receive.

It is encouraging when you are first learning to pronounce a language. When you think you got the phonemes right, then you say the words/phrases to the dragon transcription process, and voila, the right words appear in text.

It will happen that Dragon gives you a different transcribed word despite you, trying numerous times with your best efforts, to say it right. If that is the case, it is time to research and figure out if the phonemes are new to you and that they do not exist in your language.

An IPA transcript of the new phonemes will be helpful too. There are sounds that I cannot hear because of the interference of my first language. Having a visual representation of the phonemes with the IPA will probably aid me in discerning the right sound.

Once you have learned the new phonemes, you can obtain the feedback from the Dragon software and check for the production of the right transcription.

By the way:
The Rocket Language courses provide the language learner a speech recognition software. You record your voice, and you get the feedback via the transcription. It is not nearly as good as Dragon, and it is limited to be used only within the allowed words in the pronunciation drill section. But it is helpful too.

[rdearman]

Dragon is miles ahead of Microsoft.

[mentecuerpo]

This morning, I received an email with the download link and the software key for my Dragon 15 Home Edition German version.

I found the store via E-bay.

I am sending a store link for those interested in Dragon German.

https://www.blitzhandel24.de

I am sure you can find the Dragon software for other languages as well.

You will need to buy the software in an online store based in a place where your target language is an official language.

If you purchase it in an English-speaking country with only English as the official language, you only get the English version.

If you purchase Dragon from another country that has your target language as an official language, then English is also included by default.

In the USA, I can only buy Dragon English. Spanish is the unofficial second language in this country, everything is translated in Spanish, even official documents from the federal government. But Dragon is English only when bought in a USA store.

Canada has two official languages, so you will have English and French Dragon versions.

[mentecuerpo]

2014: The potential of Automatic Speech Recognition for fostering pronunciation learners'autonomy.

https://lib.dr.iastate.edu/cgi/viewcont ... ontext=etd

Even with knowledge of which sound is appropriate and a trained ear, students are still left needing to learn how to control their vocal apparatuses to accurately produce sounds. This requires practice and feedback. While there is potential for such practice and learning to be autonomous, the task may be daunting or overwhelming to students, especially those not familiar with tools or strategies that can help them.

Another technology that shows great promise for pronunciation self-access work is Automatic Speech Recognition (ASR), which allows students to experiment with the language in a safe, private setting: “ [ASR] is an independent, machine-based process of decoding and transcribing oral speech.

A typical ASR system receives acoustic input from the speaker through a microphone, analyzes it using some pattern, model or algorithm, and produces an output, usually in the form of a text” (Levis & Suvorov, 2014, p. 1). Many people now have had experience with an ASR program, either through automated telephone lines, Siri on the iPhone, or speech dictation programs. ASR is also built into many language learning programs such as Rosetta Stone (2013), Tell Me More (Auralog, 2013), and Burlington English (2014).


When used for pronunciation training, ASR is a tool that allows students to practice at their own speed, getting feedback from the words recognized. Although ASR has been criticized in dictation programs for low rates of accurate recognition for non-native speakers of the
language (Coniam, 1999; Derwing, Munro, & Carbonaro, 2000), ASR systems have been improving in evaluation accuracy for non-native speakers.

[aaronJohn1]

Wanted to quickly jump in here, I worked on a transcription platform called https://maestrasuite.com/audio-to-text that automatically transcribes audio to text. It's an online cloud based editor, unlike dragon. The transcripts can be accessed from anywhere since it is cloud based, and can be emailed and shared through public links. Would be interested to hear your thoughts!

[Querneus]

I have a professional one I use for podcasts. Can't remember the name of it right now. But the problem is more likely to be my half American, half British, half Irish accent than the microphone.

It sounds like you acquired a Boston Brahmin accent without even trying to.