Last Modified: April 3, 2009

History of My Research

The final goal of our research is "Practical realization of innate capability for acquisition of language and knowledge of a person on engineering." Concretely, we have proposed four applications using Natural Language Processing Using Inductive Learning (IL-NLP) for four kinds of data as mentioned below. Fist, we use a Japanese Kanji-Kana mixed sentence. This sentence is one kind of a character string. We proposed a morphological analysis method using inductive learning. Second, we use a pair of a Japanese Kanji-Kana mixed sentence and its Kana sentence expressed its pronunciation. They are two kinds of character strings.@ They are the same on syntax, but different on the surface expression. We proposed a Japanese Kana-Kanji translation method using inductive learning. Third, we use a pair of a source sentence and its target sentence in machine translation. In these sentences, they are the same on semantics, but different on syntax. We proposed a method of the machine translation using inductive learning with genetic algorithms. Fourth, we use a speech on spoken dialogue. The speech consists of a pair of a user's utterance and the system's reply for it. In these utterances, they are the same on syntax, but they are different on semantics. However, there is the relation of cause and effect between them on semantics. We propose a method of spoken dialogue processing using inductive learning with genetic algorithms.

Next step, we research on emotion processing in order to acquire language. Its reason is that human beings never tell anything if he/she does not have emotion. Therefore a system needs to have emotion in order to have language at first. Automatic acquisition of emotion is better. However it is too difficult now. Therefore our proposed methods can recognize emotion expressions in utterances and text.

Moreover the system can understand language and recognize emotion in order to have the sense of self. The issue of automatic acquisition of it is essential issue. Therefore we try to research on this issue in future.


Back to Homepage