Dear friends,
I have read an academic research paper about expert finder using dataset below:
Boosting (B) 46
Natural language (NL) 41
Computervision (CV) 176
Neural networks (NN) 103
Cryptography (C) 148
Ontology (O) 47
Data mining (DM) 318
Planning (P) 23
Information extraction (IE) 20
Semantic web (SW) 326
Intelligent agents (IA) 30
Support vector machines (SVM) 85
Machine learning (ML) 34
The authors show the experiments result in p@5, p@10 and p@20. I can understand if p@5 > 90%. But, I still do not undertand how they can get p@10>90% and p@20>90%. They used k-folding with k=4, it means to compute p@20>90 the test dataset consisting at least 18 items so the total dataset at least 72 (4x18). The dataset with bold typeface above are less than 72. Please explain me how it can happend?
I have sent email to the first author but has no answer yet.
best regards,
Article Learning to Rank Academic Experts in the DBLP dataset
- Badges
- Science topic