TriplePlay Plus!, Your Way, and Let's Talk all feature innovative speech recognition technology, licensed from Dragon Systems from Newton, MA, and adapted by SLS to the challenge of learning a foreign language.
Since introducing products with speech recognition, we've had many questions about the technology: What is the purpose of speech recogniton? How does it work? How well does it work? Are there different types of speech recognition? Are certain microphones recommended (other than the one that is included with TriplePlay Plus and Your Way)? Here are some answers to these and other frequently asked questions
In TriplePlay Plus!, Your Way, and Let's Talk speech recognition provides a unique, effective way to learn to speak a new language. Using the microphone, you'll speak the foreign language as you play a game or participate in a conversation. Your spoken response will be evaluated by the software to determine if you've said the correct word or phrase, and pronounced it correctly. This new method of multimedia language learning emulates the natural way a language is learned--through practice in speaking as well as in listening. The speech recognition feature will help you to improve your retention of new words and phrases, and help you to develop proper pronunciation. Perhaps, most important of all, speaking in the new language, and knowing if you are saying words correctly, will give you confidence in your ability to communicate in the new language.
How does speech recognition work?
The speech recognition in TriplePlay Plus!, Your Way, and Let's Talk does not understand speech in the sense that a human listener does. Instead, the program differentiates between or recognizes certain spoken words, phrases, and sentences in a particular language. To make this speech recognition capability a valuable language learning tool, SLS speech recognition titles have been designed with the following characteristics:
The speech recognition used is speaker independent, so it can be used by any number of different speakers without having to train the software to recognize each individual voice.
The speech recognition will recognize a beginner's accent, and within limits, a slower-than-normal rate of speech will also be recognized.
The speech recognition is a software-based solution, requiring no additional hardware other than a good-quality microphone (provided with Your Way, TriplePlay Plus!).
How well does speech recognition work?
While it is no replacement for the recognition skill of the human ear, the speech recognition used in TriplePlay Plus!, Your Way, and Let's Talk can still be an effective tool in multimedia language learning. Speech recognition is a statistical process, in which your voice is compared to target voice models that are embedded in the software. The effectiveness of the recognition depends on several factors, including the word or phrases to be recognized, the microphone volume level, microphone position, microphone quality, ambient noise level, and how well the learner matches
the intonation and rate of speech of the native speaker in the software.
With longer phrases and sentences, you can sometimes "fool" the computer by a mispronunciation or even by using some word substitutions or omissions, and you might want to try this for fun. But keep in mind that the learning is most effective when you try to imitate the voices of the speakers on the CD as closely as possible.
In some cases, trouble getting recognized is actually an audio-board adjustment problem.Technical Support at SLS can in most cases solve this problem over the phone.
Speech Recognition? Record-and-Playback? What's the difference?
The Speech Recognition feature in SLS titles allows you to use your voice (instead of the mouse) to interact--speaking a word, a phrase, or a complete sentence-- as you converse or play. The software then compares your voice to native speakers' voices and actually gives you feedback on your pronunciation.
Record-and-Playback allows you to record your voice--as you speak a word or conversational phrase--and then replay and listen to the recording of your voice. This permits you to make the comparison between your voice and the native speaker's voice.
Both are excellent tools for language learning.
Are there different types of speech recognition, and if so, how do they differ?
There are many different types of speech recognition, but a type that is used effectivelyin one application may work poorly in another. When looking at different types of speech recognition, it is important to differentiate between discrete speech and discrete speech recognition systems, likewise, between continuous speech and true continuous speech recognition systems.
Discrete speech is speech in which the words are separated by slight pauses. In continuous speech, the way we normally speak, words are not separated. Under the right conditions, and in limited contexts, a discrete speech recognition system can be used to " recognize" continuous speech. And a continuous speech recognition system might do quite well with discrete speech, perhaps even better than when applied to continuous speech.
Recognition systems for "key word spotting," which look for the presence of one or more of a limited set of key words or phrases, while ignoring the other spoken language within a sentence or phrase, can often do quite well when applied to continuous speech, but do not perform true continuous speech recognition. For example, a telephone company key word spotting system trained to spot the phrases "credit card" and "operator" might understand (correctly interpret) the sentence "I want the operator in order to make a credit card call," however, it might make the same interpretation when given "I don't have a credit card, operator." True continuous speech
recognition would be expected to differentiate between the two sentences by analyzing all words spoken and the word order (syntax).
Do I have to use the microphone that comes with the product?
You need a microphone to use the speech recognition feature, but you don't have to only use the one supplied with the product. Most high-quality directional microphones will work well with SLS speech recognition if they are matched to the microphone-input characteristics of your computer's audio board. (Unfortunately, the audio board manufacturers have not yet standardized the microphone-input characteristics adequately, but progress is being made in that direction.) And if your computer is not in a noisy environment, the microphone need not be "directional." However, some users may have a very low quality microphone that came with the audio board or
with another software product. A poor microphone might affect speech recognition
performance, as might a better microphone that is not matched to the audio board. A poor microphone may also yield muffled or tinny speech when used for record/playback. For these reasons, some of our products come with a high-quality, unidirectional, dynamic microphone that matches almost all audio boards on the market.
What is the difference between a directional and a non-directional microphone?
A microphone that is not directional is sometimes called "omnidirectional" and, for the same general quality, may be a little cheaper than a "directional" (or "unidirectional" or "cardioid") microphone. With an omnidirectional microphone, you can speak into the side or the end with approximately the same effect. A unidirectional microphone, however, should be pointed at the mouth; try to avoid speaking into the side of the microphone.