Research Article

Relation Extraction Based on Fusion Dependency Parsing from Chinese EMRs

Table 4

Performance on relation extraction for different features.

Relation typesBFsBFs + EFsBFs + EFs + DSFs1
PRF1PRF1PRF1

TrID52.7661.6756.8759.7568.3263.7564.3472.0967.99
TrWD49.3151.3150.2956.7960.5758.6261.0965.8363.37
TrCD0.000.000.000.000.000.000.000.000.00
TrAD68.5764.9766.7271.1972.7671.9776.9377.9377.43
TrNAD0.000.000.000.000.000.000.000.000.00
TrIS59.1368.1263.3164.4171.4367.7468.0176.8372.15
TrWS51.9759.3255.4055.4662.8258.9162.2467.4264.73
TrCS69.1373.9771.4772.7386.7279.1177.0390.4383.19
TrAS58.3461.4559.8562.6164.8663.7266.7868.9267.83
TrNAS0.000.000.000.000.000.000.000.000.00
TeRD73.2177.6175.3577.8980.7579.2982.8984.7983.83
TeCD59.3261.7560.5163.0964.0963.5968.9869.7669.37
TeRS81.7889.3185.3884.9893.7389.1486.9292.6789.70
TeAS61.7962.4562.1265.8967.6366.7571.8972.8372.36
DCS58.7162.7560.6661.7268.5364.9565.2772.4668.68
SID75.8278.9777.3680.6281.4181.0184.8185.4285.11
Total63.0767.2065.0767.4772.5969.9472.0976.7274.33

BFs represent basic features, extended features are represented as EFs, and DSFs are used to represent dependency syntactic features.