Aug
28
2011
A few years ago after we produced the prototype for language recognition using only stored patterns, a friend of ours suggested that the only thing scientists and engineers care about with regards to language is the accurate meaning of words. He said that is where the action is with Web 3.0 and other semantic technologies.
So like any good developer with a new requirement, we spent a couple of days and added in the additional pattern elements needed to disambiguate word senses from within the words provided in a sentence.
Disambiguation Definition
Disambiguation is the work needed to remove ambiguity - to find the correct meaning of a word from the many choices available and without the effort needed to do it using statistical approaches. This problem was in fact one of the reasons the American's abandoned their work on machine translation back in the 1960s. Much of the hard work needed to disambiguate has been done at Princeton University with their WordNet project. The hard work is compiling a list of word senses which can then be disambiguated.
For example, the word 'blue' in 'the blue woman' may mean she is sad, or she is actually the colour blue. Without the context, we will never know for sure. By applying patom theory, we can link together word senses and use our pattern matching engine to pass through only those meanings that have been experienced before as a starting point, which seems pretty good compared to the alternatives.
The following video shows an example in which sentences with colors and qualifying words like 'hot', 'cool' and 'cold' apply specifically to each other in meaning. Other examples are on the web site such as 'running water' and 'grilled bass' which provide sufficient clarity to remove most of the alternative meanings.
The key to science is understanding in order to find solutions. We feel that this approach will outperform the other approaches and provide the basis for a number of useful applications.