No one has yet solved the NL problem, and one of the reasons is that NL is characterized by a lot of rules and irregularities.
Mark's proposal is to have "really good KR" and "domain knowledge". That I don't dispute. But why not go a step further, by making the entire NL module *rule-based*, and let internet users encode the rules? The rationale is that NL is so complex that a few people will not be able to program it. Only a massive collaborative job will do. Evidence: modern computer English grammars run into 10s to 100s of Mbs. The whole problem of reference resolution (eg pronouns) can be handled by rules. Also, preposition resolution can be handled by rules, with domain knowledge specified within the rules. Comments? YKY ----- This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=fabd7936
