The virtual assistant is designed to optimize the process of holding regional elections. The functions of the finished product include collecting complaints and informing citizens on the stated topic.
A group of students from the Department of Linguistics and Translation, Sociology, and Department of International Relations, Political Science and Regional Studies of South Ural State University have developed a chatbot for election monitoring. The robot is used in the field of control and management of the electoral process, allows voters to quickly access information about the place and time of voting, available voting methods, data on candidates and other information, and also allows election control bodies to promptly respond to possible violations. In addition, the functionality of the chatbot allows for surveys of the population, obtaining data for sociological research related to the electoral process.
The first stage in creating the chatbot was the preparation of basic blocks of information. To create a list of potential questions from voters, students conducted a sociological survey. During the process, the characteristics of the target audience were studied and the problems of previous years that had been arising both at the stage of the election campaign and during the elections were taken into account. To fill the chatbot with the main source of information about the upcoming elections, voting rules, candidates, polling stations and other aspects of the election procedure, the researchers chose the official website of the Central Election Commission of the Russian Federation.
The researchers developed the chatbot’s operating scenario using linguistic analysis and worked with language material.
“The information provided to users about the election campaign should be brief and understandable,” says Elizaveta Glazko, student of the SUSU Department of Linguistics and Translation. “To ensure both these conditions, we used intent analysis of data obtained during a focus group survey, media analysis on the topic of the chatbot and analysis of information from the official website of the Central Election Commission. These procedures made it possible to identify intents – key topics of interest to voters, including available voting methods, rules of conduct at a polling station, information on who can participate in the voting, etc., as well as to establish the necessary and sufficient amount of information that corresponds to each such topic and allows to satisfy the information needs of the chatbot user. The information obtained this way allowed linguists to further work on creating a flowchart – a block diagram reflecting the logic of the dialogue development and the presentation of information blocks to users. The development of the flowchart was iterative. The ongoing discourse analysis of the dialogue text built on the next iteration made it possible to identify semantic gaps and eliminate them, ensuring, on the one hand, the coherence and integrity of the dialogue scenario in the chatbot, and on the other hand, the brevity and information capacity of the texts of questions and answers. This approach allows us to build an interaction scheme that ensures successful communication with the chatbot.”
For now, the chatbot is not capable of independently and arbitrarily changing its messages. However, it can be updated by developers, which allows it to serve different election campaigns by providing up-to-date information about the current election. This is confirmed by a number of campaigns where the chatbot was used. For example, the students’ development was implemented and made available to voters in both regional elections and the 2024 Russian presidential election. In addition, collecting feedback from users and adding information to the chatbot database that satisfies new information needs of users ensures the adaptability of the system.
The chatbot by the Chelyabinsk students includes two sections: receiving complaints from voters about various violations related to the electoral process, and an information block of broad-themed content.
“The architecture of the resource is built on the principle of transition from the general to the particular,” says Anastasia Efremova, student of the SUSU Department of Linguistics and Translation. “For example, the “Report a Violation” section consists of blocks by types of violations, each of which provides their brief descriptions; next to it is a step-by-step form for filing a complaint. All user requests are sent to the corporate e-mail of an employee of the Public Chamber, which controls the election process, for their further verification. It is important that the chatbot’s block diagram does not use complex cycles and free transitions between the “branches” of the tree simulating the development of the dialogue. Such a structure meets the requirements of the target audience and is the most convenient to manage.”
The combination of all the above mentioned methods allowed to adapt information texts of the chatbot for comfortable interaction with them by the user. To “revive” the dialogue with the bot, the students used the results of a focus group survey (eight people, aged 18 to 45), who described the degree of correctness of perception of information in the chatbot. The researchers conducted preliminary testing of the new electronic assistant with the focus group in order to correct technical, visual and textual shortcomings.
The dialogue itself in the chatbot information block is implemented using regular text buttons, and navigation when working with the bot is carried out using inline buttons. With this design, only the questions asked and the answers to them remain in the user’s dialogue with the chatbot, so at any time you can go back and read the answer to a specific question without losing the logic of communication with the chatbot.
The operation of the appeals collection block differs from the operation of the information block: it uses a finite state machine algorithm - this approach allows for control over the correct sequence of user actions in the chat. In the appeals block, data is collected from citizens using a built-in module that allows sending letters automatically. Thus, upon completion of the appeal, the chatbot asks the user to confirm sending the letter. An important task when creating the appeals block was to ensure the anonymity of users. For this reason, the students used one e-mail address from which they could send an appeal. Each appeal is assigned a unique code. The appeal code and the response code to this appeal, posted on the Public Chamber website, are the same. Thus, the collection of appeals remains completely anonymous.
Despite the fact that two functional blocks of the chatbot (informing and collecting appeals) are subject to different logic, for efficiency and optimization of the database the bot logic is stored in the form of a tree-like hierarchical structure, which represents a full-fledged chatbot script.
Such a development has first been created at SUSU in Russia and used for servicing real election campaigns. The specifics of the bot, which distinguishes it from similar resources, is its strict and verified dialogue logic, optimized language design of the bot’s questions and answers, as well as a feedback module that allows real-time monitoring of the situation at polling stations, ensuring timely response to violations and thereby increasing the efficiency of the electoral process.
Further development of the Chelyabinsk students’ research is related to improving the functionality of the system, expanding its capabilities and adapting it to specific regions and countries. They are also planning to introduce artificial intelligence and natural language processing methods to ensure a free dialogue with the chatbot and increase the productivity of its work.