Little evaluation has been performed on the many modeling options for span-based approaches. This paper investigates performances of a wide range span and context representation methods their combinations with focus scientific named entity recognition (science NER). While some most common classical encodings combination prove to be effective, few conclusions can derived representations.

How can we describe and measure real robotic sys-tems? Implemented robotic systems for intelligent tasks are are often extremely complicated, depending not only onìnteresting' theoretical parameters and models, but on many assumptions and constants which may be set almost arbitrarily. In order to evaluate and compare systems which employ diverse components and methods, we are obliged to employ ...