How to create a speaking robot

Who hasn’t read those science fiction novels where humans live next to speaking robots that can think and feel, so that they basically become better than us, because they don’t tire and they don’t die and it marks the ending era of humans who become organic batteries or slaves, or flee to other planets, or withdraw to underground cities where they plan their grand revolt…

Ok, let’s take it easy for a while. So, how do you teach a robot to speak?

Language Acquisition Device

So far humans are the only beings that have learned to communicate in language. According to Noam Chomsky (Rules and representations, 1980), humans are equipped with a built-in language processing mechanism that he named Language Acquisition Device. LAD is universal for any natural language, and genetically attached to all the human brains. It is the innate hardware, which means that we instinctively know how to process the linguistic input in order to construct rules, to understand linguistic behavior of others and to generate appropriate linguistic performance.

Universal Grammar

For LAD to be applicable to any language in the world there must be some universal rules to each language. Chomsky called it the Universal Grammar, a set of principles, some relevant, some irrelevant to a given language. Children are naturally endowed with an evaluation metric that allows them to pick the relevant parameters for their native language. For instance, there is the Pro-Drop Parameter, which determines whether a language can drop the subject pronoun, as Polish does, e.g.: Znam dobrze tego pana (I know that man well). There is no need to use “I”. In: Ja znam dobrze tego pana, “Ja” is redundant as the verb already carries information about the subject.

Language selects from a fixed set of parameters, which means that learning a new language is nothing more than changing the plus and minus values of parameters of Universal Grammar, from those of one language to those of the other. Now, a Polish student of English has to start using the Pro-Drop Parameter, to ‘switch it on’ in order to communicate well in English. And a student of German needs to activate a filter that will control the end of phrase position of a verb. And a robot would probably feel most comfortable with something like that:

“A formal grammar G = (N, Σ, P, S) is context-sensitive if all rules in P are of the form

αAβ → αγβ

where A ∈ N (i.e., A is a single nonterminal), α,β ∈ (N U Σ;)* (i.e., α and β are strings of nonterminals and terminals) and γ ∈ (N U Σ;)+ (i.e., γ is a nonempty string of nonterminals and terminals).”


And who would say that linguistics belongs to humanistic studies!

Follow up:

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.