I have implemented the Earley parser that is described in "Speech and Language Processing", including the feature structure handlers and the semantics extension. This means I can now parse simple sentences. I have deliberately kept the number of syntactic rules to a minimum, because I think I will proceed in a horizontal manner from now on. That means that I will try to parse a new sentence, process it,  and answer questions about the sentence: the full loop. This way I can create a semantic structure that fits the needs of the agent.

Speaking of which: the semantic representation of the parsed sentences is the next puzzle I will need to solve. I have implemented the semantic function of the parser that builds semantic structures along with syntactic structures. But, whereas I do have syntactic grammar rules (at least for English), I do not have semantic grammar rules. And what's more, I need rules that form semantic net structures, not predicate logic structures.