link research topics
Over the past 20 years, scientists around the world have continued making
incremental gains, and the occasional breakthrough, to improve on a range of
technologies which, when combined, can produce startling results.
Over this period, researchers at Edinburgh and Stanford have proven themselves
to be at the forefront of this exciting new technology.
The Link is involved directly or indirectly with every major aspect of language
technology research, including the following broad areas:
Speech recognition
This technology involves recognising spoken language and transforming it into
text. Technologies range from extremely accurate continuous dictation by
systems trained to recognise an individual user’s voice and accent… through to
systems that work in specific domains with background noise and must recognise
speech from any user.
Speech synthesis
By breaking down human speech into all its key phonetic parts, scientists have
been able to recreate human voices that have regional and local accents. These
so called ‘synthesised voices’ are now almost indistinguishable from real voices,
and the Link is now focussing on research involving prosody and emotion.
Spoken dialogue systems
The latest advances in dialogue systems now allow users to have a more ‘natural’
dialogue with a computer or 'software agent', rather than have to learn set
phrases. State-of-the-art dialogue technology can be used to teach as a tutoring
system, and can also enable computers to react to unexpected events in an
apparently intelligent way. Scientists are now working on spoken group dialogue
systems that involve numerous agents.
Information extraction & entity recognition
Moving from syntax to semantics is the primary goal of the latest information
extraction (IE) and entity recognition (ER) technologies. Computers can be trained
to automatically recognise entities in huge quantities of unstructured data, such as
applicant’s names and skills from a resume, and then work out the various
relationships between data sets, before finally storing the information in specially
indexed databases. This technology lays the groundwork for the vision behind the
Semantic Web, Web2.0 and intelligent search engines.
Demo. 1 |
Demo. 2
Question and answering
Advances in IE and ER, and natural language understanding, combined with
intelligent databases where data can be stored by meaning, now means
researchers can build natural question and answer systems that require little or
no training to use. These can take the form of advanced tutoring systems or
intelligent search engines that search documents according to meaning.
Summarisation
Taking a step beyond entity recognition, Link-developed technologies now enable
computers to not only interpret the meaning of phrases and paragraphs, but also
accurately summarise the content. This has proven particularly successful when
applied to structured documents like House of Lords legal arguments, and
research continues into summarisation technologies for less structured
documents.
View Applications
|
© 2005 Edinburgh-Stanford Link. |
|
