NLP for Children

Natural language processing has always been my favorite topic from artificial intelligence field. Let's remember that many anthropologists think that our elaborate language is the feature that makes us humans and distinguish us from the rest of the animals. Also, the famous Turing test for artificial intelligence says that a computer should be able to have such an elaborate conversation with a human, that the person would not be able to tell that he is talking in fact with a computer. For years the NLP field developed as computational linguistics with all kinds of grammars and tree structures, but the breakthrough is this field occurred when statistical methods were applied. On the other hand, there is a lot of research on how the children actually learn to speak. It is amazing that in about an year or so a child is able to understand what is said around her and she starts talking. In this paper are presented three cases of learning of various language features by children and the first case is how the babies of 6-7 months old learn to segment the continuous speech into distinct words. The findings suggest that they in fact apply statistical learning and segment the words on consecutive syllables the are unlikely to occur one after the other. This is just amazing, the same principle is used in many NLP algorithms. So clearly I must step up my skills in statistics to keep up with these youngsters that use them from the first months of life :)