a group-wide, corporate memory ecosystem
Video CoMem@enviaM
Our most recent CoMem video showing our cooperation with enviaM can be found here CoMem – KI zur Assistenz der täglichen Arbeit im Unternehmen (on YouTube - in German) or CoMem – AI to assist daily work in companies (on YouTube - with English subtitles)
(Intro by Prof. Dengel, then vision video by enviaM, then demo by Dr. Maus)

Corresponding flyer: CoMem used at enviaM

in English and German

CoMem@enviaM in the DFKI News 2/2022

in German and English

CoMem@enviaM at the German AI Map (Plattform Lernende System KI Landkarte) German AI Map (Plattform Lernende System KI Landkarte).

CoMem@enviaM example in one slide:
Older Showcase showing the ideas - with old prototypical UI (2016)
The short Semantic Desktop overview video from 2016 shows how our colleague Heiko Maus benefits from the interaction with the CoMem of our department. Integrations into office applications and the user's desktop show how the user's common work is enhanced but not changed.
Meanwhile we changed the UI completly - see the CoMem@enviaM example above.
See also: ForgetIT Personal Preservation Pilot
Conceptual Research
Knowledge Graph Construction: Supporting Excel in RML (ESWC 2021) Paper Teaser: The RDF Mapping Language (RML) enables, among other formats, the mapping of tabular data as Comma-Separated Values (CSV) files to RDF graphs. Unfortunately, the widely used spreadsheet format is currently neglected by its specification and well-known implementations. Therefore, we extended one of the tools which is RML Mapper to support Microsoft Excel spreadsheet files and demonstrate its capabilities in an interactive online demo. Our approach allows to access various meta data of spreadsheet cells in typical RML maps. Some experimental features for more specific use cases are also provided. The implementation code is publicly available in a GitHub fork.
See also: paper on arxiv and demo
Data Sprout demo: Confidentiality hinders the publication of authentic, labeled datasets of personal and enterprise data, although they could be useful for evaluating knowledge graph construction approaches in industrial scenarios. Therefore, our plan is to synthetically generate such data in a way that it appears as authentic as possible. Based on our assumption that knowledge workers have certain habits when they produce or manage data, generation patterns could be discovered which can be utilized by data generators to imitate real datasets. We derived generation patterns found in real spreadsheets from industry and demonstrate a suitable generator called Data Sprout that is able to reproduce them.
See also: paper on arxiv and demo
Context Spaces is our vision towards a self-(re)organizing Semantic Desktop: Tangible UIs to work with human contexts allow automatic contextual reorganization (e.g. of files and bookmarks in the video).
See also: cSpaces demo website
Spreadsheets are widely used by knowledge workers. Their methodology enables a well understood, easy and fast possibility to enter data. We present a tool which allows domain experts to use the spreadsheet metaphor to fill a knowledge base. The entered table data is then transferred into knowledge representation formats, e.g. RDF/S.
See also: RDF Spreadsheet Editor Demo
The video gives a short presentation about some of our first Managed Forgetting features (developed in the ForgetIT project, 2013-2016), explored in daily practice.
See also: ForgetIT project website
Technical Experiments
In order to provide appropriate support measures, a system like CoMem needs to constantly analyze user activities, often available as text snippets (e.g. event logs). Thus, an information extraction task, e.g. ontology-based named entity recognition, needs to be performed in (near) real-time, in order to instantly react on users' activities not to interrupt their train of thoughts. Our paper additionally focusing on inflection tolerance (not covered in the video) received the LDK 2019 Best Research Paper award.
See also: OBNER demo website
For for quick look-up tasks or Semantic Web novices, the SPARQL language is often too complex. We present a general purpose approach to convert any given SPARQL endpoint into a simple to use REST API rendering the datasource as an interlinked view of nested JSON objects that can be traversed by the API path.
See also: SPARQL Rest API Demo
DeepLinker is a tool which generates and interprets deep links to desktop resources, thus enabling references to a certain locations within a file using a simple hyperlinks. The service generates HTML representations of typical office files (doc, ppt, etc.) and allows references to document parts via URLs (URIs), so, relations can be represented in RDF.
See also: Deep Linker Demo
In feature-rich software a wide range of functionality is spread across UI components such as various menus, dialog windows, tool bars etc. Therefore, the approach of a Universal GUI Search Engine exploiting accessibility interaces was explored and first results were published in a paper that received the ICAART 2017 Best Paper Award (Area: Artificial Intelligence).
A glimpse into technical details of an algorithm for finding top-k shortest paths in RDF graphs with a new method based on Eppstein's classic algorithm. The algorithm won the ESWC 2016 Top-K Shortest Paths in Large Typed RDF Graphs Challenge.
See also: Algorithm Demo Homepage