Speech recognition
Speech recognition technologies allow computers equipped with a source of sound input, such as a microphone, to interpret human speech, e.g. for transcription or as an alternative method of interacting with a computer.
Such systems can be classified as to whether
- they require the user to "train" the system to recognise their own particular speech patterns or not,
- whether the system is trained for one user only or is speaker independent,
- whether the system can recognise continuous speech or requires users to break up their speech into discrete words,
- whether the system is intended for clear speech material, or is designed to operate on distorted transfer channels (e.g. cellular phones) and possibly background noise or other speaker talking simultaneously.
- and whether the vocabulary the system recognises is small (in the order of tens or at most hundreds of words), or large (thousands of words).
Speaker dependent systems requiring a short amount of training can (as of 2001) capture continuous speech with a large vocabulary at normal pace with an accuracy of about 98% (getting two words in one hundred wrong) if operated under optimal conditions, and different systems that require no training can recognize a small number of words (for instance, the ten digits of the decimal system) as spoken by most English speakers. Such systems are popular for routing incoming phone calls to their destinations in large organisations.
Commercial systems for speech recognition have been available off-the-shelf since the 1990s. However, it is interesting to note that despite the apparent success of the technology, few people use such speech recognition systems on their desktop computers. However, the use of speech recognition in telephone applications, for appplications like travel booking and information, financial account information, and directory assistance has been increasing as the cost for implementing such voice-activated systems has dropped.
It appears that most computer users can create and edit documents more quickly with a conventional keyboard, despite the fact that most people are able to speak considerably faster than they can type. Using both keyboard and speech recognition simultaneously, however, can in some cases be more efficient than using any one of these inputs alone. Additionally, heavy use of the speech organs results in vocal loading. Also, the typical office environment with a high amplitude of background speechs are among the most adverse environment for current speech recognition technologies.
Large-vocabulary systems with speaker-independence and/or are designed to operate within an adverse environment, however, have significantly lower recognition rates. The typical achievable recognition rate (2003) for large-vocabulary speaker-indenependent are about 80%-90% for clear environment, and can be as low as 50% for scenarios like cellular phone with background noise.
Some of the key technical problems in speech recognition are that:
- Inter-speaker differences and also intra-speaker variations are often large and difficult to account for. It is not clear which characteristics of speech are speaker-independent.
- Speech recognition system are based on simplified stochastic models, that do not match the real speech accurately.
- The interpretation of many phonemes, words and phrases are context sensitive. For example, phonemes are often shorter in long words than in short words. Words have different meanings in different sentences, e.g. "Philip lies" could be interpreted either as Philip being a liar, or that Philip is lying on a bed.
- Intonation and speech timbre can completely change the correct interpretation of a word or sentence, e.g. "Go!", "Go?" and "Go." can clearly be recognised by a human, but not so easily by a computer.
- Words and sentences can have several valid interpretations such that the speaker leaves the choice of the correct one to the listener.
- Written language may need punctuation according to strict rules that are not strongly present in speech, and are difficult to infer without knowing the meaning (commas, ending of sentences, quotations).
The "understanding" of the meaning of spoken words is regarded by some as a separate field, that of natural language understanding. However, there are many examples of sentences that sound the same, but can only be disambiguated by an appeal to context: one famous T-shirt worn by Apple Computer researchers stated,
- I helped Apple wreck a nice beach,
which, when spoken, sounds like I helped Apple recognize speech.
A general solution of many of the above problems effectively requires human knowledge and experience, and would thus require advanced pattern recognition and artificial intelligence technologies to be implemented on a computer. In particular, statistical language models are often employed for disambiguation and improvement of the recognition accuracies.
For foreign speakers an unintended side-effect of using speech recognition technology is that they can improve their pronunciation while trying to make the computer understand what they're saying.
