Abstract:
The actions of people and various infrastructural factors are the sources of events that characterize the state of the human life activity environment. The reaction of ci...Show MoreMetadata
Abstract:
The actions of people and various infrastructural factors are the sources of events that characterize the state of the human life activity environment. The reaction of citizens to such events can carry a semantic load, reflecting the state of affairs and the necessary measures to prevent the development of negative consequences. The analysis of this kind of social response can actually be the basis for one of the aspects of monitoring urban development. Such a source of information requires a study of its specificity. The main form of social response in the modern information space is news feeds of various information resources, because they contain tools for social response to ongoing events in the form of comments and emoji marking. This article describes a study aimed at developing an approach to generating a news feed using heterogeneous sources of information. Approaches to the creation of synthetic news messages that characterize the state of the urban environment and the events taking place in it, based on the parsing of real news, changes in geographic data in real news, text generation using the GPT-3 neural network, text generation based on the templating library and generation of game news based on UrbanStreetWars application are considered. The general principle of forming a single news feed based on a combination of real and synthetic news flow, as well as an algorithm for generating an integrated news flow, is proposed. The results of work the software solution based on the proposed approach are demonstrated. The issue of producing and analyzing the reaction of the urban community to the generated synthetic news flow is considered in the discussion. The results of this study will be useful in decision support systems to analyze the social response to various options for ongoing urban transformation policies.
Date of Conference: 27-28 October 2022
Date Added to IEEE Xplore: 11 January 2023
ISBN Information: