Frequently Asked Questions
Frequently Asked Questions
- What do I need to use TYWI?
- What languages does it support?
- How do I choose the best speech recognition for my voice?
- How do I choose the best automated translation for my business?
- How do I move our conference online to make it a global webinar?
- What creates optimal results?
- What if it makes a mistake?
- How good is the automated translation?
- Can I use human simultaneous interpretation not automated?
- What if we have multiple speakers?
- What if the speaker has an accent?
- What feedback should I expect?
- What is it like working with speech recognition, is it like telephone apps?
- How does the interpreter or parrot hear the original audio?
- What can hurt the results?
- What is the TYWI Mobile for hotels?
- What is the TYWI Mobile for conferences?
- How does TYWI function with Dragon by Nuance?
- What about fast speakers?
- Question-answers from a facility
- Do we provide the parrot or interpreter or do you? How does this work?
To use the text-only features like cross-language chat, you need access to the internet and an updated browser. For TYWI speech (voice) translation, you need: (1) access to the internet (2) an updated browser (3) speech recognition (included on PCs, Apple, Android, smartphones and tablets). TYWI on laptop, desktop, and the new PC tablets functions sensationally with "Dragon Naturally Speaking" downloaded from Nuance.com and installed (not the freebie app). The 3 best browsers for software are: Firefox, Chrome, and Internet Explorer. Which browser you use depends which speech recognition you plan to use. Firefox and Chrome both function well with everything including cloud recognition (Nuance NDEV) built into TYWI plus Dragon and Microsoft Speech. Internet Explorer is good with the free speech recognition included with Windows and with Dragon (by Nuance.com).
To understand how languages function, divide languages between "output" and "input".
Output: TYWI outputs 78 languages as text and 35 as synthesized voice (similar to Apple's Siri). So, for example, if someone speaks English into TYWI, the audience can choose to read subtitles (text) in 78 languages simultaneously. The same crowd could choose to hear a translated synthesized voice in 35 languages simultaneously. Or combo of both. Output languages include: Afrikaans, Albanian, Arabic, Armenian, Azeri, Azerbaijani, Basque, Belarusian, Bengali, Bosnian, Bulgarian, Catalan, Chinese Simplified, Chinese Traditional, Croatian, Czech, Danish, Dutch, English, Esperanto, Estonian, Filipino, Finnish, French, Galician, Georgian, German, Greek, Gujarati, Haitian Creole, Hebrew, Hindi, Hmong, Hungarian, Icelandic, Indonesian, Irish, Italian, Japanese, Javanese, Kannada, Khmer, Korean, Lao, Latin, Latvian, Lithuanian, Macedonian, Malay, Maltese, Marathi, Norwegian, Persian, Polish, Portuguese Brazil, Portuguese Europe, Romanian, Russian, Serbian, Slovak, Slovenian, Spanish, Swahili, Swedish, Tamil, Telugu, Thai, Turkish, Ukrainian, Urdu, Vietnamese, Welsh, Yiddish.
Input as text: TYWI supports 78 languages as typed text input (same languages as above).
Input as voice: For voice input, TYWI supports any language that is recognized by any speech recognition software that is either already installed on a device (like Windows, Apple, Android, Google, smartphones, tablets), that can be installed by the user (like Nuance.com's Dragon), or is included in the 22+ languages built into TYWI as "cloud speech recognition" (NDEV by Nuance.com). So, TYWI connects to generic speech engines for the following languages: English (many accents), French (many accents), Spanish (many accents), Arabic (many accents), Afrikaans, Hindi, Indonesian, Malay, Catalan, Croatian, Czech, German, Basque, Galician, Croatian, Bosnian, Zulu, Icelandic, Italian, Italian, Hungarian, Dutch, Norwegian, Polish, Portuguese, Romanian, Slovenian, Finnish, Swedish, Turkish, Bulgarian, Russian, Serbian, Korean, Mandarin Simplified (Beijing), Mandarin Traditional (Taiwan), Cantonese, Japanese.
The highest quality speech is Dragon (by Nuance.com) that covers: English, Spanish, French, German, Italian, Dutch, and Japanese. But any speech engine that is compatible with the Web functions. Google is excellent and covers languages that others do not, but Google will only permit a couple of sentences before you much pause significantly for processing, or it may skip sentences while it "catches up". Microsoft Speech is exceptionally fast and accurate if you have a perfectly generic accent and speak very clearly.
That's a good question. The good news is that TYWI offers a wide variety of automated translation software from which you can choose just by using a pulldown selector. Each seems to have its advantages and varies wildly by your conversation, the professional words you use, and the amount of slang or expressions you use that may not be "translatable". TYWI has set some defaults for certain languages based upon feedback, but the default may not be the best for your subject matter. The best way to choose a translation software is to run some tests with a native speaker. Takes about 3 paragraphs from a previous conversation about your business and repeat speaking those paragraphs while your linguist tries first one translation option then another. It is important to remember that you can add new words and translations (or specify that you do not want something translated) in your TYWI personal dictionary depending upon the account type you have with TYWI.
If your device is not recognizing your words well because you have an accent, try Dragon by www.Nuance.com that is spectacular at recognizing highly accented voices with amazing accuracy in certain languages. Includes various native language accents (like New York accent or Southern accent) and also non-native speaker accents, example: people speaking English with Spanish accent, or Indian, or Russian accent. Download Dragon from the non-USA website such as www.Nuance.es or www.Nuance.fr
This is surprisingly easy. Decide whether you want "audio only" or "camera". For audio, just plug a cable from the sound mixer into a separate laptop and use Skype, WebEx, or similar. Quickly stated, you just need to find a way for that audio or camera feed to go up onto the internet. It is not complicated, but if you would like us to organize the translation for you (automated or professional), then we need to know what conference software you use and the specifics of what you want to do. Please send us a Contact Us message with more info.
With this system, you can achieve perfection, achieve very good results, or achieve medium results. The "perfection" means human simultaneous interpreters (yours or ours), and they can interpret using TYWI software, online. The result is amazing and is "United Nations style", in real-time. With the automated voice translation, optimum results are achieved through 3 things: (1) good speech recognition, (2) speakers who create a voice profile (about 20 minutes), and (3) automated translation that is pre-prepared for the vocabulary. So, what is the best speech recognition for your purposes and how do you improve automated translation for your vocabulary? If you do most of your talking on the telephone, the speech recognition in Android is better than iPhone if you have any kind of accent. But if you do your speaking on a laptop with headset, Windows' speech rec is extremely good if you have a clear, generic broadcast accent and there is a $19 download that you can use to improve when Windows does not understand you. Dragon by Nuance.com is the "king" of speech recognition software; it costs $100-$200 and is available in only about 8 languages, but it understands almost everyone including people with accents, and it is very easy to improve results. Then the quality of automated translation is improved in two ways: how you speak and your personal dictionaries. If you speak clearly, in short sentences, try to have a subject and verb, avoid slang and take a very short breath between sentences, your translation will probably be very understandable. Some languages are almost perfect for certain subjects. And if you use the personal dictionaries we provide and the speech dictionaries, the accuracy increases.
Automation will not be totally perfect, but the interesting phenomena is that the listener's eyes and ears begin to jump over discrepancies and see and hear what the speaker is trying to communicate. Some discrepancies may be cute, so just laugh. Joy is part of the fun of global communication. The whole world has been using automated translation for years, and everyone wants to communicate, especially in business. In the global world, people write emails to each other full of typos because they are working outside of their native language, and people use Google Translate on websites to convert it to their language. TYWI is about communication - not perfection. It is LIVE in real-time, and it works. People understand. Some languages are more perfect than others. Absolutely. Some speakers talk more clearly and are therefore better understood. If you need perfect, use the feature for human simultaneous interpretation instead of automation. With TYWI, there is something for everyone.
In many languages the automatic translation is amazingly good. Other languages are less amazing, but all are basically understandable - the viewers/listeners understand what is happening, understand the basic conversation - anywhere from "totally understand everything" to "getting the gist of the conversation". But any way you look at it, there is communication. If you are speaking about general info, business, low tech, and basic medical, then it is quite understandable. When you begin talking about something that is very industry specific, refine the results by using the dictionary features built into both the speech recognition and into the automated translation selections offered by TYWI. Try a test first. If needs more help, prepare the software for your terminology using TYWI's dictionary feature. The more advanced speech recognition software like Dragon and Windows and TYWI's built-in "Cloud Recognition" have ways for you to correct misinterpretations. And TYWI has ways for you to enter your preferred translations of words and expressions.
Yes, absolutely. The TYWI interface provides the capability for as many languages as you desire of interpretation. The interpreters can be located anywhere in the world and can be your interpreters or TYWI interpreters. TYWI partners with companies that offer sentence-by-sentence interpreters (called "consecutive interpreters"), who arrive in 60 seconds and are paid by the minute. Our you can use simultaneous interpretation (like the United Nations) organized by your staff or using the TYWI professionals. The interpreters listen to your conference via telephone, VoIP, or the Web (e.g.. WebEx, Skype, or other), and they interpret into their headsets. What they say streams to anyone who has chosen to listen on their own devices or the AV team can send through channel earphones. There will be a little time delay. For people watching your event online is only about 2 or 3 seconds delay - so very good timing. If people in the on-site audience are listening (in the same room as the speaker), their delay is about 6 seconds because the sound goes to the web and then comes back again. But the cost savings are an advantage, and everyone in your audience can listen to a different language or read subtitles, as they choose.
The first solution is a long term solution. Each speaker can make a "voice profile" sometime in advance. The profile tells TYWI how the person speaks and pronounces in their language. Then, the AV team goes to a TYWI webpage that lists all speakers. When each person speaks, the AV person clicks on the name of that person so that the appropriate voice profile is pulled up by the software. Once a speaker has a voice profile, every time they speak in future, their voice can generate subtitles or translated voice. Speakers can also talk in their native language, which makes many people very happy.
Another option that is a bit more complex is to have one person who is the “parrot”. A “parrot” is someone who repeats every word after all speakers, talking quietly into a headset and our software. No one hears the parrot, only our software. This one person trains the speech recognition software for his/her voice. Then, his/her voice becomes subtitles, and the subtitles turn into computerized voice.
TYWI is really good for people with accents. There are 2 ways to approach the issue. If the voice of the accented speaker is being recognized by the speech recognition software, we most highly recommend using Dragon by Nuance.com (not the freebie version). It has amazing - amazing! - recognition capabilities for accented speech. Remember that you need to train the software for any voice, and that takes about 10 minutes, but the results are amazing. TYWI also provides Cloud Recognition that can also be trained in 10 minutes. It displays subtitles a bit more slowly than Dragon, but requires no download. If, however, the speaker is a guest or prefers not to use speech recognition, a parrot can repeat after the speaker in clearer pronunciation so that is well recognized by the software.
When people need language help, we usually receive feedback like 'Wow!' or 'Incredible!' or at least "I understood what they were trying to say". Use yourself as an example. Try to imagine someone speaking to you in Chinese or Japanese, and you understand nothing. Then suddenly with TYWI you understand their conversation. Overall, people who need language help are generally grateful. People who do not need language help will spot imperfection more quickly. Someone who speaks both the original language and the translated language may tend to say 'There is an error in the second subtitle' or 'The translation of X really should be Y'. TYWI was created for courtesy and inclusion - to respectfully include people in a conversation who do not speak a language and to include the deaf in everything. The translations can become more and more perfect by using professional automated translation software - also provided by TYWI - and devoting serious preparation to automation.
There are big differences between this approach to speech and the kind of speech software that you encounter when you call a big telephone system. The bank's software doesn't know who you are or how you pronounce your words so it makes mistakes. When you use TYWI's Cloud Recognition or if you install Dragon, there is a way to "train" the software to understand you. That means your results become better and better.
Any way you can get the audio to the interpreter or parrot and it is loud and clear will be fine. Telephone and Skype work. WebEx, GoToMeeting, Adobe Connect - all such conferencing software work well. The TYWI Presentation Center comes with 2 built-in interpreter channels; you just download a tiny free software (that will not hurt your computer), following some instructions for one-time setup, and you're on a roll. Contact Translate Your World to schedule a consultation on the various options.
There are a few things that can hurt the results. If you watch out for them and work around, you should be fine: (1) Not taking a few minutes to teach your speech software how you talk. It does not take long, but makes a huge difference. (2) Not pronouncing your consonants. When we speak without thinking we tend to slur our words. If you speak more clearly not only will you make a good impression on your audience but the software will understand you better. (3) Accent and unclear speaking. If you know that you have an accent or do not speak clearly and you try your speech recognition and it is not working as well as you want, try Dragon. Dragon is the "king of speech" for a reason. Often people who are misunderstood on other software will be understood with Dragon. (4) Poor environment. Some speech rec functions better in a quiet place, others are not as sensitive. (5) Try a decent microphone - not the most expensive, but not poor quality. You'll know quickly when you make some tests which works best for you.
TYWI-Hotels offers a number of really cool features for across-language speaking and texting. TYWI enables hotels to speak or text with guests wherever they are - before they arrive, during their stay, if they are lost in the city, and even after they leave. The hotel can share a personal webpage only with each guest with translated menus from hotel restaurants, local attractions, advertisements from local businesses, etc. Basically whatever you want it to be.
TYWI for conferences is a communication software between the organizers and the exhibitors and attendees plus an advertising mechanism. The complete conference agenda can be included, plus changes in schedules, biographies, and ways to set up meetings with exhibitors. It can be used to show subtitles for product demos, breakout sessions, and keynote speeches. Plus there is an advertising arm for pre-conference advertising, during conference, and post-conference including local eateries and advertising.
Personally, we like it very much, and highly encourage its use if it handles your native language. The other option is our cloud recognition that is NDEV, a variation on Dragon in the cloud. Note that the PC version of Dragon is superior to the Mac version, and the freebie app does not function with TYWI. Dragon is around $100 - $199 and has outstanding recognition, including for accented speakers. Dragon only requires a few minutes to train and it is easy to improve the recognition if it makes mistakes. The way Dragon works, it "thinks" as you speak and tries to pick the correct spelling etc. of the words you speak - for example: "there", "their", and "they're". It tends to think in full sentences and check the grammar to the best of its ability. So, it usually creates subtitles and audio that are phrases or sentences. This "thinking" does mean that it sometimes creates tiny delays while it processes, but depending upon your needs, the results are usually reliable and businesslike.
Fast speakers are an issue no matter whether you use human interpretation or automated. For automated voice translation, if the speaker's voice is going directly to the software, the speaker may test the various types of speech recognition to find which is best for them personally. Our guess is that Dragon, by Nuance, will have the best results for this fast speaker if is available in the speaker's language. If the fast speaker's voice is not the voice being recognized (the software is recognizing either the voice of a "parrot" or the voice of a simultaneous interpreter), then that person will need to condense some of the words - removing unnecessary words. Fast speakers are always an issue.
- How is TYWI used in a webinar?
TYWI is usually used as a second webpage that contains only the translation and is resized to the lower 1/4 of the screen and superimposed over the presentation webpage (like subtitles at a movie). TYWI also has an API that enables developers to use the subtitles and computerized voice translations as they prefer. Because TYWI is a "web page" that means that dozens or hundreds can click to view at the same time, and each person can select their own language. Our software will remember what language each person requested, and also provide the audio translation if their select it. Alternatively, you can choose one language and display subtitles on a large monitor or wall.
- Can we use TYWI on one computer to output to multiple languages simultaneously?
It is designed for exactly this. The speaker talks into speech recognition in a variety of ways, and as many people as you desire read subtitles or hear a computerized voice on their computers or mobile devices in their own language. Or you can throw subtitles onto the wall.
- Are we going to be sending the outputs to your servers? We worry about security.
TYWI is flotsam on the internet. TYWI handles around 10 to 20 words at a time. This breaks up any content into unintelligible pieces entering the internet independently with the stream of billions of other data. Files are a greater danger to security - complete blocks of text. This is not available through TYWI. So, if your company does not choose the "save to database" feature in our software, then Translate Your World is simply a pass-through with no record. Our software only controls the translation software selection, keeps track of who is using the system, the language selected, and preferences.
- Can we have a customized version for our requirements?
Yes, because we are the developers of TYWI, we can also customize.
The interpreter can be yours or ours. A parrot needs to be someone who speaks clearly and who can, for example, repeat the words they hear on a television show into our software. For highest quality results, either the interpreter or the parrot should be someone who is willing to spend an hour or two learning how to use the software and becoming comfortable. Our software is professional grade, so not “plug and play” at their level of usage -- and the professional approach is what improves the quality of the results. For the interpreter, it will take them about an hour the first time to set up and learn, then in future they will just double-click to join. We offer Parrot and interpreter training webinars. We can also help the AV team and conference organizers by helping with a test run in advance during which you learn how to maximize the volume of the speaker's audio signal. That is your sole responsibility — make sure that the voice of the speaker is as loud as possible so that the interpreter hears clearly. The interpreter or Parrot can be anywhere – in any country. They will listen to the event via the Web. The interpreter or parrot will speak into his/her headset, and the voice can be heard through the interpreter channel by the audiences and can become subtitles in multiple languages or even a translated computerized voice.