The problem facing speech recognition technology is 2000/12/19, the voice recognition is a complex technology, especially for Chinese speech recognition. Therefore, although many companies and research institutions have started voice technology, including speech recognition and speech synthesis technology, until the last year, in China should say this year, it has begun to have scale. There have been many large-scale speech recognition IVR systems or call centers in China, but in China, although there is a trend in combining speech recognition and call center, there is still not much application. Although there are many influencing factors, the important point is that Chinese speech recognition technology is not perfect. So, speech recognition technology should enter large-scale commercial use, what obstacles will be across? ■ User's independence so-called user's independence is that voice recognition software can identify users with different voice and accents without having to identify a particular user's sound without the need to make it through training software. Many of the current speech recognition software are identified based on standard pronunciation. In fact, people speak a thousand words, the pronunciation is different, especially for the voice of accent, but also a severe challenge for voice recognition software. Most of the current speech recognition software is through self-learning functions, constant training to reach software for specific user speech recognition. In the future, a development trend of speech recognition software is to enhance processing, which can identify various pronunciations within a certain range. In fact, there are now many voice software companies to work in this area, try to pronounce all possible dialects to income speech. In this way, the software can identify the pronunciation of the user without the special special training. ■ Natural language capability Many speech recognition systems also have natural language skills, which is the ability of software understanding the speaker. This ability is not only manifested on a specific word, but also in the phrase and complete sentences. Natural language means two points: First, it allows users to express mean by sentence. For example, in order to express checking account, users can use natural sentences with more words, such as I'd Like My Checking Account Balance. The second means that the user can use the sentence containing polysemous words. For example, we can say that I Want To Transfer $ 500 Dollars from My Savings Account To My Checking Account. This is a major standard for natural language, and users can use multiple polysemous words in a single sentence. A small number of complex speech recognition systems use a tutter dialog. Here, the speech recognition system prompts the user to answer with a word or mark. According to the survey of specialized agencies, most people prefer to use the conventional natural language, and do not like to use the prompt dialogue. People hate are imperfect natural language, when you can use a perfect natural language, people certainly irresistiblely choosing it. ■ The ability to process the new insertion of the insertion is another major advancement of speech recognition software. The inserted ability allows the user to interrupt the system when prompted, but the system can still know the user's request. This is quite meaningful for practical applications. Because people are talking, they always think about consciously thinking, often in the continuity of the language, and insert some complementary languages. Such languages are often incorrect in syntax, and routine speech recognition systems will be difficult to handle these speech. ■ Software authentication capability speech recognition vendor has recently introduced the ability of software authentication, ie, according to user voice and language characteristics, to recognize the user's ability. This is a very valuable feature in practice, as available in personnel management.
