Table 2.
For each of the 2 corpora, 10 semantic spaces were induced. | |||||||
RI spaces |
RI_20 |
RI_2 |
|
RI_4 |
|
RI_8 |
|
RP spaces |
|
RP_2 |
RP_2_sw |
RP_4 |
RP_4_sw |
RP_8 |
RP_8_sw |
The induced semantic spaces were combined in 10 different combinations. | |||||||
Combinations |
|
|
|
|
|
|
|
Identical window size |
RI_2, RP_2 |
RI_4, RP_4 |
RI_8, RP_8 |
||||
Identical window size, stop words |
RI_2, RP_2_sw |
RI_4, RP_4_sw |
RI_8, RP_8_sw |
||||
Large window size |
RI_20, RP_2 |
RI_20, RP_4 |
|
||||
Large window size, stop words |
RI_20, RP_2_sw |
RI_20, RP_4_sw |
|
||||
For each combination, 3 combination strategies were evaluated. | |||||||
Combination strategies | RI ⊂ RP30 | RP ⊂ RI30 | RI + RP |
For each of the two corpora and the conjoint corpus, 30 different combinations were evaluated. The configurations are described according to the following pattern: model_windowSize. For RP, sw means that stop words are retained in the semantic space. For instance, model_20 means a window size of 10+10 was used.