Summarizing news messages or legal text can be helpful for many users that need to process large amounts of information in limited time. Summaries should be provided at query time and hence need to be generated in real time. 

My recent research interest in summarization has focused on fast and accurate query-based multi-document summarization.  Instead of exploring computationally expensive features such as parsing, I am interested in simple but efficient features for ML-based approaches to summarization.

In addition, I see a need to improving well established automatic metrics in summarization (e.g. ROUGE) via more sophisticated NLP techniques such as parsing and semantic similarity.

 Temporal information extraction 

A mostly neglected dimension of information processing is its temporal quality. Temporal information is expressed in news messages, narratives or legal text. Extracting events and facts and possibly inferencing with them is only possible if their temporal status can be determined.

I'm interested in extracting temporal information from different text types (e.g. news, legal, medical) and reasoning with it. My  current work focusses on the integration of rule-based and machine learning approaches to extracting temporal information from text.