Under CopyrightMadan, SumitSumitMadanSzostak, JustynaJustynaSzostakDörpinghaus, JensJensDörpinghausHoeng, JuliaJuliaHoengFluck, JulianeJulianeFluck2022-03-1326.6.20182017https://publica.fraunhofer.de/handle/publica/40072910.24406/publica-fhg-400729Biological signaling is complex and our knowledge about it is often only available in literature. Signaling can involve small molecules as well as proteins that can be activated or deactivated by various regulations such as ligand binding, complex forming, modification status or miRNA binding. Changes in signaling influence biological processes and/or are involved in disease etiology. The Biological Expression Language (BEL) has been created to store this kind of information in a structured form that can be used for network generation and visualization as well as interpretation of experimentally generated data. The BioCreative VI BEL track provides training data and an evaluation environment to encourage the text mining community to tackle the automatic extraction of such complex relations hips as well as converting it to BEL. Although only a few groups participated in this track, the groups participating the second time could drastically increase their performance. The best system reached 32% F-score for extraction of complete BEL statements (task 1) and, when given the named entities, above 49%. Beside rule-based systems, methods involving hierarchical sequence labeling and neural networks are adapted to this task. For the second task in the BEL track, finding evidence text snippets for a given statement, despite the provided training data, only one team took part.en003005006518conference paper