Table 9.
MSH WSD results at different k-NN threshold levels
| AEC | MRD | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| 66 | 75 | 85 | 95 | 66 | 75 | 85 | 95 | ||
| LSC | 0.8370 | 0.8371 | 0.8371 | 0.8377 | 0.8071 | 0.8070 | 0.8071 | 0.8071 | |
| SG | Coll | 0.8173 | 0.8214 | 0.8268 | 0.8327 | 0.8082 | 0.8077 | 0.8073 | 0.8071 |
| CollParser | 0.8284 | 0.8271 | 0.8337 | 0.8355 | 0.8076 | 0.8071 | 0.8071 | 0.8071 | |
| LSC | 0.8391 | 0.8413 | 0.8416 | 0.8400 | 0.8072 | 0.8072 | 0.8072 | 0.8071 | |
| kAEC | Coll | 0.8252 | 0.8331 | 0.8385 | 0.8407 | 0.8104 | 0.8104 | 0.8100 | 0.8092 |
| CollParser | 0.8298 | 0.8337 | 0.8396 | 0.8409 | 0.8098 | 0.8093 | 0.8090 | 0.8090 | |
Disambiguation results in terms of accuracy using the MSH WSD set. Several k-NN values are used in combination with the semantic group (SG) and the automatic extracted corpus (kAEC) methods. The disambiguation methods AEC and MRD are compared.