|
Speech Recognition for Smart Homes4.4 Open-ended commands
Imagine a computer-based butler. This would need to be able to interpret, understand, and
carry out complex commands. For example 'please book me a night flight to Tokyo next
Thursday and reserve a suitable suite at the Grand Hyatt Hotel', or perhaps, 'please check
my address book and make a dental appointment at an appropriate time over the next few
days'. There is not only a significant vocabulary size implied in such commands, but also the
understanding of the meaning of the command, and an appreciation of related factors: for
example the words 'suitable' and 'appropriate' demand value judgments.
Quite clearly these commands impinge on the area of natural language processing (NLP)
research, but do represent the ultimate destination of research into smart home systems –
the home which is able to conveniently cater to our needs. It is likely to be several years
before such open-ended commands can be handled successfully by automated systems.
4.5 General ASR applications
Since the advent of computers, the need to collect and neatly present documents has
required textual data entry. The potential of ASR systems is that much of this process can be
performed through VI. Many data entry applications involve predefined items such as name
and address, for example form completion, package sorting, equipment maintenance, and
traffic accident reports.
Data entry applications usually have limited vocabulary size including numbers, name and
address details, and several additional control words. However, there may also be a
requirement for a significant number of application-specific words, depending on the
application type (McTear, 2004).
Similarly, computer games will likely form potential future applications which can only be
guessed at currently. Educational use also had great potential – much teaching is conveyed
vocally in schools and universities worldwide, and an individual education for children,
adapting to their pace and needs, may well become a reality with advanced VI systems.
Finally, most societies contain people battling with loneliness, nobody to talk to, nobody
who understands them. The thought of a machine companion may seem far-fetched today,
but in the opinion of the authors it is only a matter of time before a sufficiently powerful and
responsive computer system could become a best friend to some in society. As with the
entire smart home project, what is needed is a sufficiently natural and accurate VI system
coupled with a computer system that could pass the Turing test, at least when conducted by
their potential clients.
|
| |