VACOS 2 - Natural language and conversational support for object search

• Applicants: Professor Dr. Norbert Fuhr; Dr. Dagmar Kern
• Subject Area: Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing ; Human Factors, Ergonomics, Human-Machine Systems
• Term: since 2018
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 388815326

The VACOS project aims to provide better support for searchers with vague information needs formulated in natural language. We consider two use cases: product search and research data search. In the first phase of the project, our human-centered development process focused on analyzing the user context, creating datasets with natural language information needs, enriching datasets with topical and natural language attributes, and developing prototypes for object search. The results of the first project phase have shown the necessity to searchers not only during the initial search process but also to explain search results and their relation to the formulated search request. Therefore, concepts of the research areas "Conversational Search" and "Explainable Search" will be used in the second project phase. First, object metadata will be enriched to generate object representations that better match information needs formulated in natural language. Based on this, mixed-initiative natural language search dialogues will be developed. For the presentation of search results, various representation forms will be investigated, focusing on vagueness and explainability. Single result items will be explained with respect to the formulated search request as well as in comparison to other found objects. In the context of this proposed project, the prototypes developed in the first project phase will be extended by new dialog and explanation components. Human-centered evaluations will be conducted to examine natural language and conversational supports for object search.

• DFG Programme: Research Grants
• Co-Investigator: Dr. Daniel Hienert