Guiding Theme A2: Identification of complex event structures in discourse

This guiding theme aims to identify event structures („who did what to whom“) and relations that connect them in discourse, as a basis for content selection in multi-document summarization (MDS). Next to detecting event structures within a single discourse, the analysis of event structures in a MDS setting can exploit cross-document techniques. Special attention will be paid to the analysis of events in heterogeneous document sources stemming from different genres and domains.

Ph.D. projects may focus on different aspects related to the event identification problem in a MDS setting.

Redundancy-oriented web-scale IE methods are severely restricted in recall on the individual document level. Within the MDS setting, we need to achieve precise and recall-oriented identification of events. Event analysis will thus build on syntactic-semantic parsing architectures. Event detection models can be enhanced through knowledge obtained from cross-document argument and event alignment. The recognition of complete event structures in discourse involves the analysis of contextual relations between them. Among these are non-locally bound semantic roles and semantic relations between events, e.g., causation, that need to be detected. Also, events tend to be realized with different degrees of granularity across documents of different genres and origin. Such granularity differences, including linguistic realization differences may be bridged using techniques from textual inference.

The guiding theme A2 will interact with A1 and A3 in Area A, and will provide important selection criteria for summarization in Area B, especially B2. Graph motifs (C1) may be investigated as features for discourse phenomena. Special learning methods may include deep learning (C3) or ranking techniques (C2).

Poster (in German)

Example thesis topics

  • Cross-document techniques for event identification in MDS
  • Modeling discourse contextual relations between events
  • Genre- and domain-adaptive event identification


Jun Araki and Teruko Mitamura (2015): Joint Event Trigger identification and Event Coreference Resolution with Structured Perceptron. Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, p. 2074-2080.

Michael Roth and Anette Frank (2012): Aligning Predicates across Monolingual Comparable Texts using Graph-based Clustering. Proceedings of the 2012 Conference on Empirical Methods in Natural Language Processign (EMNLP), 2012. 

Michael Roth and Mirella Lapata (2015). Context-aware Frame-Semantic Role Labeling. Transactions of the Association for Computational Linguistics, vol. 3, p. 449-460.

Quang Xuan Do, Yee Seng Chan and Dan Roth (2011): Minimally Supervised Event Causality Identification, Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing, p. 294–303.



A A A | Drucken Print | Impressum Impressum | Sitemap Sitemap | Suche Search | Kontakt Contact | Webseitenanalyse: Mehr Informationen
zum Seitenanfangzum Seitenanfang