Multilingual Context-Aware Chatbots for Multi-domain Test Data Generation

Various novel communication channels have been developed in recent years. Examples are messaging apps, social media platforms, and packet-oriented telephony services. This development is accompanied by an increasing need for methods and tools that extract and explore the large amount of natural language data for forensic analyses. However, these tools have to be configured, trained, and evaluated before they can be deployed. For forensic test data, the provision of high-quality domain-specific data in various formats is of utmost importance. In this chapter, we present a chatbot framework for generating high-quality domain-specific forensic data sets for chat and audio communication. While actual chatbots are mainly implemented as human-to-bot systems, we have developed a bot-to-bot framework for interactive data generation. Our framework is open-source and based on state-of-the-art natural language processing methods. It allows a transfer of the synthetic generated data in real-world data types and different languages. We evaluate the quality of the generated data and discuss their usability.