Assessment of the emotional tonality of the text: analysis of problems and solutions

Abstract

The article examines modern approaches to the analysis of the emotional tonality of a text — one of the key tasks of natural language processing. Special attention is paid to various areas of tonality analysis, as well as to a number of problems faced by one or another approach. The article discusses the basic concepts characterizing the analysis of emotional tonality. In particular, the directions of the subject areas in which the considered analysis tools are used are highlighted: the market sphere, medicine, politics, social network analysis, sports, etc. At the same time, the analysis of tonality is clearly highlighted in relation to similar areas of research. The levels of work with data in relation to the subject of analysis are also defined, including documentary, supply level, aspect analysis, as well as multi-domain and multimodal approaches. Each described area of analysis of emotional tonality is considered in the context of modern works devoted to a particular approach. At the same time, current tasks and possible ways to solve them are mentioned. The article also discusses the problems of cross-domain knowledge transfer, denial management, sarcasm recognition and elimination of lexical ambiguity. In conclusion, the importance of further research is emphasized, among which special attention is paid to the direction of interlanguage analysis, the expansion of linguistic resources for languages belonging to representatives with small or medium-sized resources, and the creation of adaptive systems capable of processing texts in various languages, taking into account their specifics. It is also noted the importance of developing large-scale data corpora and methods to eliminate algorithm bias.