Computer-To-Computer Based Communication Through Natural Language

Abstract

While highly accurate and efficient computer-to-computer communication exists, the exploration of communication of computer models via natural language is still worth exploring. This paper explores the creation of a “Speaker” model which preforms image-to-text operations and a “Listener” model which does the reverse text-to-image task. These models can be used together to form the basis of computer natural language communication not only in existing languages such as English but in completely new generated languages as well with the help of the “Rambler” model which combines the Speaker and Listener to preform the entire image-to-text-to-image process. By comparing the image on both sides of the process, how effective the communication is can be measured. While natural language-based computer communications will likely never be common place, it nonetheless poses some interesting and unique challenges.