Abstract
Trying to combine speech recognition and natural language understanding to create a system capable of understanding conversational dialogues, we face a number of problems that are not encountered in text processing. Oral speech, as a rule, is more concise, less grammatically correct, less structured and more ambiguous than text, it is characterized by the absence of smooth fluency, processes of assimilation and reduction, divergence of syntactic and prosodic articulation of the speech sequence, temporal characteristics such as the length of pauses, the presence of hesitation pauses. , duration of speech segments, melodic characteristics. The interest of phoneticians is directed to the search for the causes of phonetic variability, to the study of the typology of the variability of sound phenomena of a general linguistic nature or specific, inherent in specific language groups. In addition, speech recognition systems that attempt to extract words from utterances typically result in word insertion, deletion, or substitution errors due to incorrect speech segmentation and perception. Prosodic means of intonation can be helpful in understanding the structure of a dialogue, as can parsing. This article presents some results of the use of prosodic means to convey semantic nuances in sounding discourse, the implementation of intonation functions and various approaches to the study of intonation ins ounding discourse.