Theoretical Research: A Computational Theory of Language Comprehension
We are currently developing a computational theory of language comprehension that could explain how children acquire their first language. The key ideas are as follows:
- The basic parsing mechanism is to combine "mental sketches" of words perceived to those already held on a stack.
- Mental Sketches are complex entities, each containing both what is traditionally known as the lexical and semantic information about a word. We make no distinction of such knowledge. A child at different stages of development will have different knowledge about a word.
- We claimed that one of the earliest knowledge about words acquired by children is the way in which each word is combined to those already perceived. For English, the method used is based solely on word ordering.
- We have developed a method using different labels of ?R's and ?L's to successfully parse English sentences containing different adverbial phrases, adjectival phrases, prepositional phrases, transitive and intransitive verbs, wh-words, questions and others.
This research is currently being funded with a government NERF grant ($1.2 million for 4 years from 2002-2006).
- Yeap W.K., Reedy P., Min K. & Ho H., Visualizing the meaning of texts, Proceedings of the Nineth International Conference on Information Visualisation, 2005, 883 - 888.
- Yeap W.K., A new GOFAI theory: How language works, Proceedings of the International LISP Conference, 2005, 371 - 378.
- Yeap W.K., Semantics parsing revisited or how a tadpole could turn into a frog, Proceedings of the 2nd Language & Technology conference, 2005, 468 - 472.
- Ho H., Min K. & Yeap W.K., Pronominal anaphora resolution using a shallow meaning representation of sentences, Proceedings of the Eighth Pacific Rim International Conference on Artificial Intelligence, 2004, 862 - 871.
- Yeap, W.K. A sketched computational theory of language comprehension, Proceedings of the Cognitive Science Conference, 1998, 1170 - 1175.
Professor Henry Huang, UCLA, California, USA
Professor Allan Bell, Director, Auckland University of Technology, New Zealand
Using the new algorithms above, we are currently developing a commercial product, SmartINFO, for the visualisation of text.