Masterarbeit
Designing a Scanner for Identifying and Analyzing Chatbot Interfaces on the Web
Completion
2026/01
Research Area
Students
Andrey Shevandrin
Advisers
Verena Traubinger M.Sc.
Dr.-Ing. Sebastian Heil
Description
Chatbots are increasingly integrated into web services, but unlike traditional graphical user interfaces, they do not yet have standardized best practices for their graphical components. Such chatbot patterns, which are design or interaction elements which are used in recognizable ways, are not yet analysed on a wide variety of chatbots on the web. A systematic approach is therefore needed to discover and document the common patterns used in chatbot implementations. This analytical challenge is also influenced by the high volatility of chatbot deployments, and the absence of comprehensive datasets of publicly available chatbots for systematic study.
To address this challenge, the proposed work centers on identifying and analysing chatbot implementations through the development of a specialized web scanner. This approach first involves a manual analysis of source code to find recurring implementation strategies which are evaluated and integrated into the scanner. Secondly, it applies web interaction techniques to activate and capture the dialogue flow. By combining web scraping with dynamic web rendering, the scanner can gather data about structural properties of chatbot implementation.
The objective of this thesis is the creation of a solution or the combination of existing approaches to solve the problem described above of identifying and analysing chatbots on the web. This comprises the following parts. An analysis of the state of the art on web chatbots, current chatbot interface research, web scrapers, automated rendering of chatbots, and other relevant work. The prototypical implementation should include the development and evaluation of different approaches for scanning web pages to find chatbots, as well as the dynamic rendering of the chatbot interface and analysing found patterns. A suitable evaluation should be conducted where the prototype is tested on benchmarks regarding its identification of chatbots, the comprehensive and accurate analyses of the chatbot interface, and their compliance with elicited requirements based on the literature research.
