Smart Data & Knowledge Services

In the topic field knowledge work, we are investigating how information and knowledge workers can be effectively supported in their work.

To illustrate this, one can imagine a personal knowledge assistant - or, to put it more bluntly, an "information butler" - who looks over the user's shoulder and identifies the information required in each case from the most diverse information sources, data silos or legacy systems of the company as well as its own data collections and proactively makes it available for the current problem. For example, the right task for an incoming e-mail, similar solutions for a new ticket, the correct inventory file for a request.

We use approaches such as the Semantic Desktop for the personal view and embedding in the personal infrastructure as well as Corporate Memories for the views of groups, departments, the company. These are based on semantic technologies that enable the generation of machine-understandable knowledge graphs from the multitude of different data sources (such as calendars, address books, documents, web pages, notes, file directories, databases, legacy systems). By embedding them in the work, rich contexts can be derived and used for storing data/information/knowledge and realizing knowledge-based services.

To this end, we are pursuing the following topics, among others

• Corporate Memory & Semantic Desktop,
• Semantic Technologies, Evolutionary Knowledge Graphs
• Neuro-symbolic approaches: combination of language models (LLM) and knowledge graphs both for building and expanding knowledge graphs and for retrieval, as well as providing facts for chatbots,
• Personal knowledge management, business process-oriented knowledge management,
• Context, information needs, proactive information delivery,
• Building personal information models (PIMO; personal knowledge graph) and corporate memories (organizational knowledge graphs) from distributed, heterogeneous information sources,
• Assistance & knowledge acquisition embedded in daily work, tasks, processes, agents,
• Dealing with small and large text collections using faceted and semantic search, entity recognition, clustering, RAG & knowledge graphs (GraphRAG), etc.,
• Digital forgetting, managed forgetting, information assessment,
• Practical application: access permissions, scalability, continuous operation, etc.
   

Our DFKI CoMem unites our research approaches and industry solutions in one ecosystem. More information: https://comem.ai

We combine research with industrial applications.  For example, we are collaborating with enviaM, where our DFKI CoMem is already in productive use, and we are working with Wacom on the ‘Semantic Ink’ approach for semantification of handwritten notes.
 

Selected projects:

• CoMem – Corporate Memory
• TransferLab Cybermapping with the German Federal Bank
• DIAL - AI solutions for the sustainable digitization of SMEs
   

Finished projects:

• SensAI – Self-organizing Personal Knowledge Assistants in Evolving Corporate Memories (BMBF)
• Managed Forgetting – Sustaining Grass-roots Organizational Memories: Foundations and Methods of Managed Forgetting for Knowledge Workers (DFG)
• Managed Forgetting II – Sustaining Grass-roots Organizational Memories: Methods and Effects of Applying Managed Forgetting in Administrative Corporate Scenarios (DFG)
• supSpaces – Semantic support knowledge spaces for agile knowledge management in the 3rd - Level Support (BMBF)
• ForgetIT – Concise Preservation by combining Managed Forgetting and Contextualized Remembering
• EPOS – Evolving Personal to Organizational Knowledge Spaces
• myRPA – Experience-based Robotic Process Automation for Knowledge-based Personal Assistants (BMBF)

Publications:

https://comem.ai/home/publications