Artificial Intelligence Research Group

Some Past Projects
Under Development

11. Using Natural Language as Domain Knowledge

Investigator(s): Mr Jeremy Coker and Professor Max Bramer

Vast corpora of natural language data are available on many diverse topics. At the same time there is widespread recognition that induction is usefully guided by domain knowledge. Unfortunately the process of acquiring this knowledge negates some of the usefulness of machine learning technology: the knowledge acquisition bottleneck is merely shifted, rather than eliminated. We seek to extract sufficient knowledge from natural language data to be useful in guiding the inductive process in associated domains. As part of this work a study looking at the use of domain knowledge and explanation of its application by human domain experts and humans naive in those domains, is being conducted. A software tool allowing users to interactively construct decision trees has been implemented, and it is hoped that the information gathered from this study will be used to derive mechanisms for applying more natural forms of domain knowledge. It is also hoped that comparisons of trees derived mechanically with and without domain knowledge with the expert derived trees will yield useful insights on the general inductive process.

Next