A Grammarless Language Generation Algorithm Based On Idiotypic Artificial Immune Networks

The immune system is capable of evolving by learning from its environment over the lifetime of the host. Using the ideas of idiotypic network theory and artificial immune systems we explore the analogy between the immune system and linguistics to suggest a new approach to build a network of sentence phrases and train it using a learning algorithm. The learning algorithm is devised to help evolve the network sufficiently with stimulations by correct phrases or antigens. The network after sufficient stimulations, suppressions and decay is capable of detecting and differentiating between correct and wrong sentences. We verify with experimental data and observe promising results for such an immune network based algorithm. The system learns a language without any grammar rules similar to a small child who knows nothing about grammar yet learns to speak in his native language fluently after a few years of training.