The future of speech recognition

4 000 000 iPhone 4S Apple only sold during the first weekend of its release. The main arguments in favor of buying, Siri - built-speaking assistant, understands asked questions and providing answers.
Siri correct operation of the system allows complex processes running in the background. Researchers and technology visionaries believe that the speech recognition solution for the future. Analyst
Shannon Cross of Cross Research's speech considered to be the most
natural medium to issue commands and is of the opinion that - thanks to
the software, Siri - Apple in a few years mastered the speech
recognition market. Even
Bill Gates believes that voice control will be successful - but only
when the command recognition accuracy will be sufficient. For this reason, Microsoft's equivalent of Siri, known as Tellme, still did not go beyond the R & D department.
5000 words in 1984 could identify the speech recognition system from IBM, and every single analysis dealt a few minutes. His modern successors can distinguish more than 300 000 words and recognize individual words takes only seconds. However,
while manufacturers boast recognition accuracy of 99 percent., It
specifically Polish language makes computers a hard time. First, the average user of the Polish language, operating an average of 4,000 words, it has an advanced vocabulary. To express a similar content of American will need only about 800 words. In
addition to the enormous difficulties associated with distinguishing
homophones, words with the same pronunciation but different in meaning.
3
seconds on average, need to understand the question asked in the game
show "Jeopardy," and give the correct answer created by the IBM computer
Watson speaking naturally. For
its creators - David Ferrucciego - right next to the most important
thing is to recognize words primarily semantic and statistical analysis.
With
these tools, devices in the future will be able to understand the
actual content of speech and search for needed information and manage
contacts and appointments.
Siri - the facts
Military
invention: In 2003, the agency DARPA (Defense Advanced Research
Projects Agency) commissioned the Institute to create an intelligent
assistant SRI called CALO (Cognitive Agent That Learns and Organises). The contract value amounted to approximately $ 175 million. In 2008, on the basis of previous work prepared for the consumer version - Siri. Apple acquired Siri project in 2010.
Other devices running iOS: Using a new trick, you can also install Siri for older iPhone 4 and iPod Touch. However, the device must first be unlocked via jailbreak procedure.
Fuel
Transfer: Service Ars Technica measured the amount of data sent from
Siri to Apple's servers while processing a single query - an average of
37 KB.
Areas of application of speech recognition
Tools notingUnlike other areas of application software for personal computers is important to adjust to a specific user. To achieve high accuracy of recognition, you need to read the voice dictating tool.
Effectiveness:
Groups of people often use the same professional vocabulary, such as
lawyers or doctors, achieve recognition accuracy of 99 percent. Writers and journalists, using a wide vocabulary, they must pay a much lower efficiency.
Automated call centersCompanies and agencies use simplified voice recognition systems, understanding only the digits or a single word.
Effectiveness: Results when choosing simple options are very good. But
even a slight accent makes it impossible to obtain information - then,
like many other call center, comes into play alive employee.
TranslationsAppy for smartphones such as the free Google Translate tool to translate languages in real time.
Effectiveness: The results are surprisingly good. This application Google recognizes several languages.
CarsMercedes,
Audi, Ford and Fiat use to control equipment and mobile multimedia
solutions leader board car market - the company Nuance. Just
like the iPhone 4S Siri, a voice-controlled navigation system is able
to recognize names from your address book connected to the phone.
Efficiency:
Depending on the intensity of noise when driving and the quality of the
microphone system is able to recognize almost all of the commands, if
they are not integrated into complex sentences.