Add ESA and LSA approach.

[chenditc]

For ESA (Using wordnet vector):

word to sense level:

1. word-word approach: Pearson's correlation 0.132543 Spearman's rho 0.212888
2. word-description approach: Pearson's correlation 0.188277 Spearman's rho 0.221309
3. description-description approach: Pearson's correlation 0.187931 Spearman's rho 0.174960

phrase to word level:

1. phrase to definition approach: Pearson's correlation 0.230682 Spearman's rho 0.234178
2. definition to definition approach: Pearson's correlation 0.224463 Spearman's rho 0.212838

[chenditc]

For LSA, the api is too confusing.

LSA stands for "latent semantic analysis" or "lexical semantic analysis"?

[chenditc]

Try wikipedia

[chenditc]

semantic vector package, need data for training

[chenditc]

For ESA (Using wikipedia vector):

word to sense level:

1. word-word approach: Pearson's correlation 0.149545 Spearman's rho 0.137075
2. word-description approach: Pearson's correlation 0.227022 Spearman's rho 0.217455
3. description-description approach: Pearson's correlation 0.177256 Spearman's rho 0.178998

phrase to word level:

1. phrase to definition approach: Pearson's correlation 0.230682 Spearman's rho 0.234178
2. definition to definition approach: Pearson's correlation 0.227466 Spearman's rho 0.232829