Figure 4
From: Gimli: open source and high-performance biomedical name recognition
Comparison of windows and conjunctions results on development sets. Comparison of F-measure results achieved by token-based and optimized windows and conjunctions in the development sets of both corpora, considering exact matching evaluation, different model orders and text parsing directions. Results for the JNLPBA corpus indicate the overall performance, i.e. across entity types. FW: Forward, and BW:Backward.