by Tim Margheim

Natural Language Parsing involves training a computer to determine, “What does this sentence mean?”  That includes determining the sense of each word.  However, if you want a system to be able to answer questions or draw conclusions, you must go further.  The system must also be able to answer, “What does this imply?”

For instance, take the sentence, “Humans eat meat.”  If the Brain knows this, how should it answer questions like:

• Do humans eat food?
• Do humans eat bacon?
• Do humans eat vegetables?
• Do primates eat meat?
• Do primates eat food?
• Does Jeffrey eat meat?
• Did Jeffrey eat that meat?

What’s the logical relationship between “meat” and specific kinds of meat?  Or between “human” and something more general, like “primates”?  Does “Humans eat meat” mean “Some humans eat some meat”, or “All humans eat some meat”, or “All humans eat all meat”?  How do you define the connections between each category, and how do you store the statements about humans, so that the implications will be clear?

And how about number?  Do you account for the difference between “I have a car,” “I have one car”, and “I have only one car”?  What does each one imply for the question, “Do you have two cars?”

Deduction requires more than storing the sense of individual words.