Project Turing
Alan Turing
was one of histories great scientists and the inspiration for much work in computing
and robotics.
Project Turing was a major Thinking Solutions project run throughout 2006 and 2007,
named in honour of the late Alan Turing. The prototype demonstrates that it is possible
to implement a new type of brain-like programming on a computer.
Goal
The project goal was to create an improved program to converse with people using
text. To do this, the complexity of grammar first needed taming into a simpler form.
While a human brain is able to recognise multiple meanings of word sounds, a computer
program isn't available to analyse the options and return the best result.
Once grammar is controlled into its core underlying elements, the communications
received and sent can be tracked and used as the information store for the brain.
Finally, once the system deals with grammar and its resulting patterns of higher-level
interaction, linking additional information will permit useful conversations. Each
element need only rely on the same approach - storing, matching and using linked
patterns.
Achievements
Project Turing has not yet achieved all its objectives, however the first (grammar
recognition) has passed its proof-of-concept milestone. Language patterns can now
be identified within other complex patterns and within a sea of ambiguity. These
same patterns are used to provide appropriate language generation capability. Indeed,
despite it being a separate requirement, the same capability enables foreign language
generation provided the appropriate word-sense is identified for the target language.
A number of lessons were learned during the early prototyping phase, which remain
incorporated in the design to allow for future capabilities within existing products.
These capabilities include:
- Translation of new phrases from the source language to the target language using
the same patterns used for matching.
- Incorporation of additional grammatical attributes from other languages - English
doesn't have as much differentiation for gender as other languages, for example.
- Additional pattern use, such as the additional word forms found in French verbs.
- Identification of "word senses".
Status
Project Turing was briefly on hold in early 2007 and again in late 2007 due to work
towards the grammar tool's proposed release in early 2008. Following the trial phrase,
feedback indicated that a number of additional features were highly desirable including:
- Inclusion of a larger dictionary, as the trial version contained around 10,000 words,
instead of a more typical English Language dictionary with 100,000 or more.
- Display of the meaning of matched words, instead of just the grammatical value matched,
to help students and machines to understand the sentence.
- Inclusion of a reduced number of displayed word meanings, based on the phrase matched.
Where phrases and clauses are matched that disallow some word meanings, do not include
them in the list. (This is known in the field as word-sense disambiguation)
- The feature to enable a structured analysis of a number of sentences, without the
need to type in each one such as to support students reviewing a set of examples.
- The ability to display more types of clause patterns, such as those from VerbNet.
Thinking Solutions is currently using its Project Turing prototype to verify the
extent to which these requests can be incorporated into the first language tool
released. We have utilised the WordNet dictionary's words (approx 150,000) to extend
our dictionary at this stage. Our target is to complete the prototype work by 1
February 2008 at which point we will update this web site with our program's new
capabilities.
In the meantime, screen shots are available in the tour showing the types of displays