WHO IS SPEAKING: MAN OR MACHINE?

Abstract

The presented paper gives a detailed survey of peculiarities of the AI-powered speech synthesis in the aspect of their lingual differences from the natural speech generation. The chosen subject matter for investigation is motivated by the rapid development of Information Technologies and modern applied linguistics in the field of human speech generation, and namely – Natural Language Processing (NLP) versus robotics (embodied AI) speech production. The ultimate objective of our investigation consists in making an attempt to establish formal and communicative differences between machine and natural speech synthesis on the material of the English dialogue carried out after the “human-robot (H-R)” model. Despite the enormous theoretical and practical contribution of linguists and computer scientists to machine language generation (Joseph Weizenbaum, Roger Schenk, William Aaron Woods, Terry Vinograd, N. Chomsky), many aspects of this problem require further consideration. The latter include morphological disambiguation of the English language, dependency grammar limitations to recognize and synthesize lengthy syntactic structures (compound and complex sentences), restrictions in the cognitive system of modern robots to identify abstract concepts and determine the so-called ad-hoc categories that people can form on the spot, grounding on the situations they face, understanding metaphorical phrases which belong to one of the most important linguistic phenomena reflecting human natural way of thinking. In a broader sense, the remaining problem is building robots that will be capable to predict the social behaviour of people, understand their physical and emotional states due to synchronic system “human-robot” (not just the physical simulation). The scientific novelty of the article consists in figuring out lingual and epistemological differences between human and machine speech. Our investigation results in concretising the notion “artificial intelligence”, establishing the main stages of AI-powered speech synthesis and foregrounding the basic differences in generating the artificial and natural English dialogue.