Figure 2.

Multi-level representation learning for ER model. B-Sym: beginning of the noun phrase for the symptom entity; B-Test: beginning of the noun phrase for the test entity; C: input sentence; E: input embedding; I-Sym: middle of the noun phrase for the symptom entity; I-Test: middle of the noun phrase for the test entity; O: not a noun phrase; Trm: transform-block; y: output sentence’s predicted tag sequence.