Different types of voice-recognition packages have different degrees of functionality: speaker-dependent vs. speaker-independent systems, command /control vs. dictation systems, and discrete vs. continuous systems.
A voice-recognition dictation system that is speaker dependent requires a user to "train" the software before it can be used. Therefore, this type of system is sometimes called a trained voice-recognition system. Training consists of reading through a manufacturer-provided list of words or sentences that contain all the different sounds in a language. Often the user is required to pronounce each word on the list several times before moving on to the next word. Altogether, the training can take several hours to complete. Once the training exercise is complete, the computer makes calculations based on the voice data that it has received. These calculations are used to draw up a voice profile that matches the speaker's vocal patterns. No two speakers sound exactly the same (we have different pitches, tones, accents), and a speaker-dependent system that has been trained by user A cannot then be used successfully by user B. In order for user B to be able to work with the system, he or she must also complete the training exercise and create his or her own voice profile.
In contrast, speaker-independent voice-recognition technology, also known as universal voice recognition, does not require user training. Typically, the manufacturers have "pre-trained" the software using speech samples from a wide range of people. Therefore, a user may begin using a speaker-independent voice-recognition program upon installation. At first glance, speaker-independent systems seem like a good option; however, if the user's voice patterns fall outside the sample range (e.g., if the user speaks with a foreign accent), it may be difficult to get the system to work properly. On the other hand, even speaker-dependent systems that have been trained to recognize the voice of a specific user may not work all the time. For example, if the user has a head cold, his or her pronunciation may temporarily change and may not be recognized by the system.
Another distinction between voice-recognition systems is whether they are command/control systems or dictation systems. Command/control systems allow users to interact with the computer by giving a limited set of commands, which usually correspond to the commands found on application menus (e.g., "Open," "Copy," "Paste"). In contrast, dictation systems allow users to enter new data into the computer by dictating instead of typing. Command/control systems cannot be used to convert a printed text into electronic form; this process requires a dictation system. Some hybrid products combine both command /control and dictation features into a single system.
Systems that allow dictation can be further broken down into discrete (or interrupted) speech vs. continuous-speech systems. With a discrete system, the user must pause between words so that the computer can distinguish where one word ends and the next begins. For most speakers, this entails modifying their normal way of speaking, reducing it to a slower and more stilted pace. Some ... users ... find ... this ... difficult ... or ... frustrating. In contrast, a continuous-speech system allows users to speak at a more normal pace, though pronunciation must still be very clear.
责任编辑：admin
上一篇：Some Different Types of Electronic Corpora
下一篇：Optical Character Recognition
微信公众号搜索“译员”关注我们，每天为您推送翻译理论和技巧，外语学习及翻译招聘信息。