Projects

Current projects

"COMUTE" is a DFG-funded project in collaboration with the Martin-Luther University of Halle/Saale and the Freie Universität Berlin. We aim at semi-automatically aligning multilingual, semi-parallel versions of text, as they can arise in different disciplines of the Humanities and the Social Sciences. COMUTE runs from 2023 to 2026.
"Transforming Text across Media" is a project in the second phase of the DFG-funded SFB (Collaborative Research Center) 1287 "Variability in Language". The Co-PI is Jun. Prof. Tatjana Scheffler (RU Bochum), and two Berlin-based companies are our application partners. We aim at semi-automatically transforming texts for optimal presentation in different written and spoken media. The project runs from 2021 to 2025.
"Predicting limits of variability in discourse using neural models" is a project in the second phase of the DFG-funded SFB (Collaborative Research Center) 1287 "Variability in Language". The Co-PI is Prof. David Schlangen. We develop neural models for distinguishing different degrees of coherence in text. The project runs from 2021 to 2025.
"Trajectories of Conflict" is funded by DFG and AHCR in a joint German/British initiative.The Co-PI is Prof. Chris Reed (Univ. of Dundee). We study linguistic mechanisms of signaling conflict and compromise, using the debates of the UN Security Council as target corpus. The project runs from 2021 to 2024.
"LARGA" is a project in the DFG SPP "RATIO: Robust Argumentation Machines", the Co-PI is Prof. Stein (Bauhaus Univ. Weimar). We aim at modeling and implementing rhetorical strategies in argumentation. The project tuns from 2021 to 2024.

Past projects

"AKILAS" (Adaptiver KI Assistent für die Schule) was a BMBF-funded project aiming at building an adaptive learning assistant for school children. The project was coordinated by our lab; partners were Prof. Böhme (Educational Science, UP), Prof. Stober (AI, OGVU Magdeburg) and Solocode GmbH (Berlin). It ran from 2021 to 2024.
"Imperceptron" was an Exist-funded initiative of three students, targeting the recognition of terms for ICD coding in medical text. It ran from 2020 to 2021.
"Discourse Strategies across Social Media: Variability in Individuals, Groups, and Channels" was a project in the DFG-funded SFB (Collaborative Research Center) 1287 "Variability in Language". It studied variation in phenomena of coreference, common ground management, and coherence relations in different social media, and it ran from 2017 to 2021.
"UPracticeML" was a BMBF-funded project on expanding the role of machine learning in our MSc program Cognitive Systems. We also collaborated with various local industry partners. The Co-PI was Sebastian Stober (Junior Research Group Leader in Cognitive Science), and the project ran from 2017 to 2020.
AnaKonn was a DFG project aiming at connective-informed shallow discourse parsing for German. We payed particular attention to the anaphoric properties of certain connectives and their influence on the task of selecting the arguments. The project ran from 2017 to 2020.
We were an active member of the COST Network Action TextLink: Structuring Discourse in Multilingual Europe (Coordinator: Prof. Liesbeth Degand, B-Louvain). It ran from 2015 to 2019.
"LIA - Learning Intelligent Agent" was an Exist Forschungstransfer project on the analysis of costumer care communication data and dialog system research funded by the BMWi and the ESF. It ran from August 2015 till August 2017.
SFB 632/D1 was part of the Potsdam/Berlin Collaborative Research Center on Information Structure (IS), funded by DFG. D1 ran throughout the SFB, from 2003 to 2015. We have developed the ANNIS linguistic database for multi-layer corpora, whose third edition ANNIS3 is freely available. The recent foci of the Potsdam branch of the project were the automatic classification of IS notions in text, and the interplay between discourse structure and IS. Since 2007, the Co-PI was Anke Lüdeling (HU Berlin).
SFB 632/T2 is part of the Potsdam/Berlin Collaborative Research Center on Information Structure, funded by DFG. T2 ran from 2011 to 2015 and aimed at a systematic integration of information structure concepts into speech synthesis technology. We collaborated with DFKI Saarbrücken (MARY speech synthesizer) and with Beyo GmbH Potsdam. The Co-PI in the project was Frank Kügler.
Discourse analysis of social media was a collaborative project in a BMBF program on e-humanities. It ran from mid-2012 to mid-2015 and involved partners in Communication Science (Prof. Neuberger, LMU Munich; Prof. Quandt, U Münster) and Business Informatics (Prof. Stieglitz, U Duisburg).
Multiple collective identities was a collaborative project in a BMBF program on e-humanities. It ran from mid-2012 to mid-2015 and involved partners in Political Science (Prof. Kantner, U Stuttgart) and Computational Linguistics (Prof. Heid, U Hildesheim; Prof. Kuhn, U Stuttgart).
Sentiment analysis was a short-term project (2012/13) with ferret go GmbH on transferring resources for sentiment analysis from German to English. The project was supported by ILB, Potsdam.
SemScribe (BMWi/AIF, 2009-2011) was a joint project with a Berlin-based company in the ZIM program. We developed a prototypical system for clinical text generation, which produces a medical report instantenously while the doctor is keying in his or her observations.
Multi-document text summarization (2008-2010) was a BMWi/AIF-funded joint project with a Potsdam company in the ZIM program, where we developed technology for summarizing multiple texts on the same topic, in particular news stories and film reviews.
Causal connectives in automatic discourse understanding (2006-2008) was a DFG-funded, joint project with IDS Mannheim (Dr. Eva Breindl). We studied properties of German causal connectives, built a lexicon for them, and used it in a prototype system for automatically finding connectives and their arguments in text.
Dealing with Uncertainty (2005-2007) was an EU-funded project in the Marie-Curie scheme. We designed a spoken dialogue system around a board game, which served as a testbed for handling uncertainty on various levels of analysis.
PINK/Text-summarization (2004-2007) was part of a BMBF-funded network on XML technologies, which involved a number of companies in Berlin and Potsdam, as well as FU Berlin. We developed a prototype system that combined statistical and linguistic methods for, partly genre-specific, automatic text summarization.
Semantic Web for Pathology (2003-2006) was a DFG-funded, joint project with FU Berlin (Prof. Robert Tolksdorf) and HU Berlin Charité Clinic (Dr. Thomas Schrader). We automatically built semantic representations of medical reports (pathology), mapped them to an ontology, and thus enabled semantics-based retrieval.
In addition, we ran several short-term collaborative projects with local companies.