Skip to Main Content

The results of the experiments applying the scenarios described in Section 6 (see Table 4) are presented in Table 6. The table shows the results of the ten scenarios set out in rows, and in columns we provide the scenario identification (column 1); the model applied out of the 32 generated, following the notation in Section 7.2 (column 2); the number of features in the original model (column 3); the number of features in the model adapted for that scenario after removing noninformed features, that is, the features used in the original model that do not fit in the description of a concrete scenario (column 4); and the accuracies of the original and final model (columns 5 and 6, respectively). In each scenario we applied the best model of the 32 we generated taking into account the features that each model uses and that fit the best in each scenario according to the hypothesized available linguistic processors. When there is no concrete model to project how the ADN-Classifier would perform in a concrete scenario, we selected the model that fits approximately in that scenario and removed the noninformed features. For instance, Scenario 10 describes the case where the nominal lexicon is not available or the nominalization candidate is a noun that does not occur in the nominal lexicon, and the features used are extracted from the parsed tree at lemma level and from the SRL process in order to obtain argument structure information. Because we do not have a model that perfectly fits in that scenario, we select the LEAFF (lemma based model using examples from the corpus as the unit of classification and obtaining the features from both lexicon and corpus, with full vocabulary and full corpus sets), and we removed all the features from the lexicon except the ones related to the argument structure, simulating an SRL process.32

Table 6

Experiment and evaluation of scenarios.

Scenario
Model
Initial Att.
Final Att.
Initial Acc. (%)
Final Acc. (%)
LELFF 1,559 1,559 85.62 85.62 
SELFF 1,671 1,671 96.65 96.65 
LEAFF 1,755 1,755 87.20 87.20 
SEAFF 1,867 1,867 95.46 95.46 
LELFF 1,559 1,416 85.62 85.56 
SELFF 1,671 1,613 96.65 96.17 
LEAFF 1,755 1,611 87.20 87.12 
SEAFF 1,867 1,808 95.46 95.41 
LECFF 197 197 84.86 84.86 
10 LEAFF 1,755 1,556 87.20 87.08 
Scenario
Model
Initial Att.
Final Att.
Initial Acc. (%)
Final Acc. (%)
LELFF 1,559 1,559 85.62 85.62 
SELFF 1,671 1,671 96.65 96.65 
LEAFF 1,755 1,755 87.20 87.20 
SEAFF 1,867 1,867 95.46 95.46 
LELFF 1,559 1,416 85.62 85.56 
SELFF 1,671 1,613 96.65 96.17 
LEAFF 1,755 1,611 87.20 87.12 
SEAFF 1,867 1,808 95.46 95.41 
LECFF 197 197 84.86 84.86 
10 LEAFF 1,755 1,556 87.20 87.08 

Close Modal

or Create an Account

Close Modal
Close Modal