Academic Lens Research Assistant (alpha)

Academic Lens is a data-driven research assistant, designed to navigate and reason through a corpus of materials around a specific research topic. The current knowledge base comprises the complete proceedings of the International Conference on Interactive Digital Storytelling (ICIDS).

The assistant responds to the researcher's questions or keyword searches, and aims to provide relevant citations for every answer. The system is trained on selected sources, derived from peer-reviewed publications as part of trusted proceedings. We expect our users to reflect on the generated responses and linked sources to supplement their personal research endeavours. Unlike other chatbots, this tool does not support creative writing or text editing tasks.

The system uses a Closed-World architecture for evidence-based enquiry: responses are based on the provided corpus. If the answer is not contained within the source materials, the system is designed to state its limitations rather than speculate. Researchers can refine the system's focus by selecting specific sources from the library for targeted analysis.

The project is currently undergoing testing, though this service will be intended for research purposes, offering convenient access to fact-checked and context-aware information for the research community.

Usage

• Entries (questions or keywords) should relate directly to the source materials listed. To maintain academic integrity, the generated responses should be supported with verified references from our database. In other words, the app will deliberately have a hard time “hallucinating” new information.

• To filter the sources used, right-click relevevant papers listed in the topic library, or those cited in the previously generated results. Select 2 or more sources.

• Sustainability is a priority for this project and our user base. The LLM model is self-hosted and doesn’t rely on GPU-based processing. Response time may be slower than with commercial AI chatbots; we are actively working to improve this. For now, consider your questions carefully before submitting.

• Sessions may be saved to your local file system as a JSON, which can then be loaded again for future reference.

• Please note that questions and responses may be logged for research and quality improvement purposes.

• This is an alpha-stage prototype; errors and inconsistencies are to be expected. Please consider using the built-in feedback to help our design and development.

Sources

This public version of this chatbot uses preprint versions of academic materials for research and information synthesis. These may differ from the final published versions. Where possible, references or links to the official published articles are provided. Users are encouraged to consult those versions for authoritative content. The generated responses are for informational purposes only and do not replace the original publications.

Contact

Have any questions? feel free to write to our team:

Project Managment

mattia.bellini [at] ut [dot] ee

Research & Development

primett [at] ut [dot] ee

Co-financed by the EU.