We can use computers to model meaning and then do things with it. In this talk I introduce very informally two complimentary approaches: the Open Multilingual Wordnet, a large lexical network of words grouped into concepts and linked by typed semantic relations and Head-Driven Phrase Structure Grammar, which models syntax and structural semantics. Both have been developed in a dynamic approach where we first build a model, then apply it to a corpus, find problems, fix them, and reapply, gradually approaching (but never quite reaching) perfection. Finally, I will talk a little about how we can use these models to analyze text for sentiment.