ARCHITECTURE OF AN INTELLIGENT CHATBOT FOR SMART CITY: AI-BASED RECEIVING, ANALYSIS AND ROUTING OF CITIZEN REQUESTS

Abstract

The paper addresses the development of an intelligent digital service for receiving, analysing and routing citizen requests within a Smart City ecosystem, in which a Telegram bot serves as the primary communication channel. The aim is to design a service architecture that combines a conversational interface, a multi-step submission scenario, storage of multimedia attachments, and an automatic thematic classification module followed by routing to responsible municipal departments.

The service is built on a finite-state machine, SMS OTP authorisation and an S3-compatible store; classification quality was compared across three models. On a corpus of 3,260 Russian-language requests, a RuBERT-class model achieved the highest quality (accuracy 0.914; F1 0.907), outperforming linear baselines. Simulation of two processing scenarios showed a reduction of the average time to forward a request to the relevant department from 26.15 to 6.85 minutes.