Keynote

Extracting Knowledge from Scientific Text

Abstract:

How to Leverage Domain Expertise and Linguistic Categories

The design of new experiments in scientific domains heavily depends on domain knowledge and on previous studies and their findings. However, the set of scientific publications that are relevant to a particular problem is typically very large and grows every day, making it hard or even impossible to keep track of all experiments conducted for a particular research question. In this talk, I will present joint work with the Bosch Center for Artificial Intelligence and domain experts that addresses this problem in the materials science domain by creating sophisticated datasets and building neural information extraction systems.

In addition to the dealing with complex domains, extracting information from scientific text is challenging because professional academic writing is far from only stating undisputed facts. Authors make ample use of hedges, linguistic devices indicating uncertainty, as scientific propositions are usually considered as valid only until they are overwritten by newer findings. Distinguishing valid solutions to problems from unverified and/or potential solutions is a crucial step in information extraction from scientific text. To this end, I will present recent corpus-linguistic work on the functions of modal verbs (can, may, should, etc.) in scientific text, and on argumentative zoning.

In the final part of my talk, I will show how transformer-based models can help with patent classification in the context of patent landscape studies.

Speaker:

Annemarie Friedrich is a Professor for Natural Language Understanding at the Institute of Computer Science at the University of Augsburg. Her research interests are computational semantics, discourse processing, and linguistic annotation. Before joining the University of Augsburg in 2023, she worked as a Senior Expert for Natural Language Processing at the Bosch Center for Artificial Intelligence, where she focused on text mining for the scientific domain. She also worked as a postdoctoral research at the Ludwig-Maximilians-Universität in München. She holds a Ph.D. in Computational Linguistics from Saarland University, for which she received an IBM Ph.D. Fellowship and the 2018 GSCL Doctoral Thesis Award.

Annemarie is the Vice President of the German Society for Computational Linguistics and Language Technology (GSCL) and this year’s program co-chair for KONVENS. She regularly serves as (Senior) Area Chair for *ACL conferences, as Action Editor for ACL Rolling Review, and reviews for ACL, NAACL, EACL, EMNLP and Coling. She is a member of the committee of the ACL Special Interest Group for Annotation (SIGANN).