Building bridges : a cognitive science approach to a human Dolphin dialog protocol

Cognitive science is the interdisciplinary study of cognition, mind and intelligence that has come out of the bridging between philosophy, psychology, artificial intelligence, neuroscience, linguistics, and anthropology. It has so far not been too concerned with cognition in an inter-species perspective. This thesis is looking at inter-species communication, or more specifically, human-dolphin communication, from a cognitive science perspective. Three aspects of language: a system, a brain, and an interface are seen as inter-dependent requirements for language acquisition. The bottlenosed dolphin meets the requirements pretty well. Apart from vocal motor-control, big brains, and a cognitive understanding of symbolic reference and syntax, bottlenosed dolphins have excellent memory, live in complex social groups, vocalize a lot and are good at mimicking. This makes them very good candidates for an attempt to investigate inter-species communication. The linguistic system is seen as a 'complex adaptive system' not in individuals, but between individuals. Frege, Saussure, Complexity theory, memetics and universial Darwinism are discussed, and incorporated in an inter-disciplinary view on language systems. Brains and intelligence is also discussed. Brains are seen as 'pattern-acquisition devices' that has evolved the ability to adapt to a changing environment. The plasticity of brains is seen as a key ingredient in this. Knowledge from neuroscience, and artificial neural networks are introduced. Some comparisons across species are being made, and methods to calculate intelligence are being discussed. Douglas Hofstadter's ideas about analogy-making being at the heart of intelligence are introduced. The need for an interface to handle the symbolic reference aspect of language is emphasized. The lack of voluntary motor control of speech organs in many species is disqualifying their participation in a language scheme based on these conditions. Bottlenosed dolphins are shown to have such voluntary motor control, but not compatible with the sounds used in human linguistic communication. This interface problem must be addressed and overcome somehow if two-way communication between humans and dolphins is ever to take place. The General Audio Pattern Recognizer (GAPR) - is a software that bridges this gap. It allows humans to make, and recognize dolphin sounds. A new kind of communications bridge is presented in this thesis. The General Audio Pattern Recognizer (GAPR) is a tool for classifying and transforming audio patterns in general, into symbolic representations, and for linking combinations of these symbols to words. GAPR can be used both for setting up a language structure, and to run an (almost) real-time translation for dialog or game. Several instances of GAPR can be linked up with each other, in a server-client relationship, using the Tcp/Ip protocol. The different instances are linked up either locally on the same machine, on separate machines in a local area network (LAN), or through the Internet. Digitizing of audio, feature extraction, classifiers, neural network architecture, Viterbi search, self-organizing feature maps, learning-vector quantization, and dynamic time-warping are being described. A bridge metaphor is used throughout the thesis, and language is seen as a process of crossing gaps at many levels. At its top level, language is a bridge between two or more individuals, enabling an exchange of intentions and ideas that, by many people, is considered the most important innovation in the history of mankind. By using the bridge metaphor we can look at language and communication issues in a more detached way without getting caught up in the particularities of human language. Inter-species communication, one of the biggest gaps around, is not likely to be bridged with a 6-lane-freeway right away. Part of the thesis is an attempt to narrow the gap intellectually, by pointing out structural similarities that will simplify the construction. The rest of the thesis will be looking at the remaining distance, and what kind of methods can be utilized to over come it. The tremendous advances in computer power and new algorithms for processing data is today making it possible to address the issue of inter-species communication in a new way, something that was not possible even few years ago. This thesis then, is both an interdisciplinary survey of language and cognition in general, and an engineering job. The purpose of the project is not to prove that dolphins can acquire language, but to make that investigation possible by providing a bridge