Although much of the work carried out within the Language Technology Group over the last 5 years has focussed on shallow processing of text, we are now beginning to explore applications where useful and interesting results can be obtained using a more traditional rule-based approach which requires a representation of meaning and knowledge. I will focus particularly on recent LTG work involving spoken dialogue systems. In general, the overall success of such systems depends on achieving an adequate understanding of the speaker's intended meaning, even if there are errors in word recognition or syntactic parsing. The kinds of questions which then arise are: * How does taxonomic reasoning interact with other kinds of inference in natural language understanding? * How does plan construction enter into the semantic interpretation of natural language commands? Ewan's home page is: http://www.informatics.ed.ac.uk/people/staff/Ewan_Klein.html