Comparison with Knowledge-based Methods


In this section we compare three different knowledge-based methods: conceptual density Agirre1996, a variant of the conceptual density algorithm david_julio2001; the Lesk method Lesk1986 ; and the specification marks method.

Table 6 shows recall results for the three methods when applied to the entire SemCor collection. Our best result achieved 39.1% recall. This is an important improvement with respect to other methods, but the results are still far below the most frequent sense heuristic. Obviously, none of the knowledge-based techniques and heuristics presented above are sufficient, in isolation, to perform accurate WSD. However, we have empirically demonstrated that a simple combination of knowledge-based heuristics can lead to improvements in the WSD process.


Table 6: Recall results using three different knowledge-based WSD methods
WSD Method Recall
SM and Voting Heuristics 0.391
UNED Method 0.313
SM with Cascade Heuristics 0.311
Lesk 0.274
Conceptual Density 0.220