Researcher Assistant for the RoHub system

The Researcher Assistant is an advanced chatbot that helps researchers use the ROHub system. The chatbot introduces an innovative approach to exploring RoHub resources through a new communication channel for users – a digital researcher assistant that guides researchers through RoHub features, stimulates exploration of its resources, and provides answers to questions and problems. Interaction with the bot is consistent, relevant and conducted in a conversational tone. “The latest version of the chatbot, based on the use of LLM, has shown great potential in helping ROHub users to get information about the platform, its main concepts and related functionalities. The new functionality will soon be implemented in the production environment after testing and optimisation”, explains Raul Palma. Marcin Wolski adds that the chatbot for RoHUB is not just another tool among many others, but a completely new solution using the latest technologies and approaches. The tool is based on the open LLama3 model and embedded in the PSNC infrastructure, thanks to which we have full control over the data flow and its security.

Selected features of the Researcher Assistant: How to use ROHub, ChatBot guides the user through the whole process of using ROHub, helping to create and manage scientific objects such as articles, research data or projects. Creating scientific objects: The wizard explains how to create a scientific object in ROHub, helps you to define the type of object (e.g. article, dataset) and fill in the required fields. Integration with search mechanisms: The assistant allows you to search for research objects in ROHub, tells you how to filter the results and customise the search criteria. Research object recommendations: The wizard analyses the user’s previous searches and identifies similar research objects.

The assistant uses cutting-edge and innovative technologies such as the Large Language Model (LLM) and Retrieval-Augmented Generation (RAG). LLMs are the core technology of the next generation of AI bots (such as ChatGPT or Google Gemini). They are trained on huge amounts of text data, such as books and articles, to learn patterns and relationships between words. RAG plays the role of a fact-checker for LLM. It allows the model to check real-time information from reliable sources before developing answers, making its answers more accurate and reliable, explains Maciej Łabędzki of the Data Engineering and Analytical Platforms Department.

If you want to create your own research path, join the RoHUB community – the Research Object Hub.

For more information: https://www.rohub.org/