Skip to Main Content

Therefore, we approach the problem of classification taking into account different feature sets which come from different knowledge resources, and we examine and evaluate the task performance when a decreasing number of knowledge resources are used. Depending on the available knowledge resources and natural language (NL) processors, we designed the classification task in different scenarios, which are presented in Table 4. The columns include the knowledge resources used in each scenario (column 2), whether the features used are extracted at sense or lemma level (column 3), and the NL processors that are necessary in each case.

Table 4

Description of the scenarios used for evaluation.

Scenario
Knowledge Resources
Features level
NL Pre-Process
1 AnCora-Nom+AnCora-Verb lemma POS 
2 AnCora-Nom+AnCora-Verb sense POS+WSD 
3 AnCora-Nom+AnCora-Verb lemma POS+Parsing 
4 AnCora-Nom+AnCora-Verb sense POS+WSD+Parsing 
5 AnCora-Nom lemma POS 
6 AnCora-Nom sense POS+WSD 
7 AnCora-Nom lemma POS+Parsing 
8 AnCora-Nom sense POS+WSD+Parsing 
9 – lemma POS+Parsing 
10 – lemma POS+Parsing+SRL 
Scenario
Knowledge Resources
Features level
NL Pre-Process
1 AnCora-Nom+AnCora-Verb lemma POS 
2 AnCora-Nom+AnCora-Verb sense POS+WSD 
3 AnCora-Nom+AnCora-Verb lemma POS+Parsing 
4 AnCora-Nom+AnCora-Verb sense POS+WSD+Parsing 
5 AnCora-Nom lemma POS 
6 AnCora-Nom sense POS+WSD 
7 AnCora-Nom lemma POS+Parsing 
8 AnCora-Nom sense POS+WSD+Parsing 
9 – lemma POS+Parsing 
10 – lemma POS+Parsing+SRL 

Close Modal

or Create an Account

Close Modal
Close Modal