Table 3.
Mean contribution of algorithms to word sense disambiguation (smoothing) and ranking
SMOOTHING ALGORITHM | TRAINING SET n = 12,056 | TEST SET n = 8,131 |
Concept | 12% | 13% |
Homonym | 1% | 1% |
N-Gram | 55% | 53% |
Metaphone | 5% | 4% |
Length | 14% | 14% |
Part-of-speech | 10% | 11% |
History | 3% | 4% |
TOTAL | 100% | 100% |