Dr. Ali Farghaly
The ultimate aim of Artificial Intelligence is to model human intelligence inside the machine. This includes modeling natural language which enables humans to communicate with each other and transfer knowledge from one person to another and from one generation to other generations. Natural language made accumulation of knowledge possible and unique to Man. Once we create smart machines that have the natural language capacity, humans can interact with the machine the same way they interact with each other i.e without learning a programming language or to how to code. The following is an account of some of the properties of natural language we need to model inside the machine.
1. Primacy Speech
Natural language is primarily spoken. Out of the 6000 wold languages, a very small number of them are written. Many have no writing script and are never written. The machine should be able to ‘listen’, ‘understand’ acoustic signals and should be able to generate appropriate vocal response.
An important property of natural language is its creativity. Humans have no problem producing and understanding novel sentences they never heard or produced before. Humans do not learn their language through memorizing or storing massive linguistic data sets. On the contrary, children acquire their native language through the exposure to limited and degenerate data. By the age of five, every normal child internalizes the grammar of his native language.
Productivity. Another property of natural language is its productivity. It is not limited to any domains. It serves its speakers by allowing them to talk about everything and anything. It lets them talk about the past, the present, and the future. They can create poetry, write novels, describe things that never existed. There is no limit to what we can say in our language.
Natural language is inherently ambiguous. It is ambiguous at the lexical, syntactic and semantic levels. However, native speakers of the language rarely feels the ambiguity because they interpret what they hear using their linguistic and real world knowledge as well as common sense.
Variability. Natural language provides its speakers with the ability to vary what they say. They can say the same thing in different ways and at different levels. They can speak formally or informally and they vary their speech automatically according to the context and the type of interlocutors they are talking to.
Unlike programming languages, Esperanto and Morris code, no one knows who invented or designed any natural language in spite of the complexity and sophistication of the grammar of any natural language. Natural languages develop, grow strong and sometimes die exactly like any organism. They have a life of their own.
Children acquire their native language by the age of five regardless of how easy or difficult their native language is. They also go through a uniform learning stages. They start by babbling, one word stage, putting lexical words together and then complete sentences.
8. Duality of Structure
All natural languages can be analyzed at two levels: the level of the meaningless phoneme (unit of sound) and the level of the meaningful morpheme (units of meaning). Any sound like /p/, /u/, or /t/ has no meaning by itself. But once we put two or more sounds together they make sense as in ‘put’. With a small number of sounds in each language. Speakers can create an infinite number of sentences about an infinite number of topics.
The challenge is how to design a smart computer with all these capabilities? Smart young brains are invited to take the challenge!